Latest News

Artificial intelligence (AI) is making significant strides in enterprise IT, moving beyond the initial phase of AI assistants that helped employees with routine tasks. The next wave of innovation focuses on AI agents that can understand operational context, make recommendations, and take action independently. This shift is particularly evident in infrastructure and IT operations, where agentic AI is becoming a key trend in managing complex environments.

The evolution from AI Assistants to Autonomous IT Operations follows a clear trajectory: AI assistants provide support for employees, while AI agents work toward defined objectives and perform multiple steps across connected systems. Enterprise applications are expected to increasingly include task-specific AI agents that move beyond productivity support to completing end-to-end tasks. For instance, in IT operations, these agents can investigate infrastructure alerts, identify potential root causes, optimize cloud resources, manage repetitive configuration tasks, recommend or execute remediation, and support capacity and performance management.

The crucial change here is that AI is transitioning from providing information to participating directly in operational workflows. This marks a significant shift toward more autonomous IT operations, where human teams are no longer solely responsible for monitoring complex environments. Modern IT environments span across cloud platforms, data centers, applications, networks, APIs, and AI services, generating vast amounts of operational data that can be difficult for humans to correlate manually.

Agentic AI can help bridge this gap by connecting signals and providing operational context. Microsoft’s work on agentic observability in 2026 is a prime example, integrating logs, metrics, traces, topology, and operational context to facilitate issue identification and accelerate investigations. Moreover, Microsoft has introduced autonomous operations in preview, where AI continuously performs preparation and triage while humans retain control over decisions that change the environment.

The objective of this new approach is not necessarily to eliminate human involvement but to reduce manual work required for responding to operational signals. This points toward a new operational cycle: observe → understand → decide → act. However, as AI agents become more active, traditional observability may no longer be sufficient. Organizations need visibility into the behavior of AI systems, including model outputs, agent actions, tool usage, performance, and interactions with enterprise systems.

IBM introduced AI Agent and LLM Observability in 2026 to improve visibility into production AI systems as agents interact with APIs, data pipelines, and enterprise services. This raises critical questions about AI observability: what did the AI agent do? Why did it make that decision? Which systems did it interact with? Did the action produce the expected result? Can the action be traced and audited?

AI observability is a foundation for trustworthy autonomous operations. However, greater autonomy also introduces new risks. An AI agent with access to infrastructure can potentially make changes faster than human operators but may also create security, availability, or cost problems if it makes incorrect decisions.

To mitigate these risks, organizations need to establish clear boundaries around what AI agents can do independently. Key controls include role-based access and permissions, human approval for high-impact actions, continuous monitoring, audit trails, policy-based controls, and testing before production deployment. The direction is toward controlled autonomy rather than unrestricted automation.

The longer-term opportunity lies in shifting from reactive IT to continuously optimized operations. Traditional operations follow a pattern of alert → investigation → decision → remediation, while AI-driven operations can shorten this cycle by allowing agents to correlate events, investigate issues, recommend actions, and perform approved remediation. However, fully autonomous IT will not happen uniformly; only a minority of AI agents are expected to become fully autonomous in the near term.

The next stage for enterprises is less about adding AI assistants to existing software and more about redesigning how technology is operated. This evolution combines AI agents with observability, automation, governance, and enterprise data. The result could be a broader shift from software that helps people operate systems toward software that can participate in operating those systems.

The organizations that benefit most will not necessarily be those that deploy the most AI but rather those that combine AI capabilities with reliable data, strong governance, and clearly defined operational boundaries.

The latest iteration of Google’s flagship smartphone series, the Pixel 11 lineup, has been officially unveiled. This new generation is designed to work seamlessly with Gemini Intelligence 1, a powerful AI engine that enables users to navigate their daily lives more efficiently. The result is an impressive array of camera upgrades and proactive assistance features that make everyday tasks easier than ever before.

The introduction of Gemini-powered voice input feature, Rambler, marks a significant improvement in voice recognition technology. This innovative tool allows for fast and accurate transcription of spoken words, even when users employ filler words like ‘um’ or ‘ah.’ The integration of Sign-to-text functionality within Gboard is another notable advancement, utilizing Google DeepMind’s SL2T model to enable signers to naturally communicate through text.

Pixel 11 devices will also provide timely and relevant information through easily accessible cards in various apps. For instance, when a user receives a message about an upcoming trip, their booking details will be displayed along with real-time updates on flight status. Additionally, the phone will offer contextually relevant suggestions to streamline tasks such as saving events to calendars or providing location-based insights.

One of the most exciting features of Pixel 11 is Magic Capture, which enables users to automatically snap photos in critical moments without having to intervene. This AI-driven capability analyzes around 400 frames per session and delivers high-quality images with perfectly timed captures. Furthermore, Magic Capture can apply edits like crop and unblur, making it easier for users to share their precious memories.

The Camera Looks feature allows users to choose from a range of aesthetic options at the point of capture, catering to individual preferences. This includes three default looks – Natural, Shadows, and Vanilla – as well as six additional styles that can be customized to suit personal tastes. The ability to create custom camera settings is another significant advantage for photographers.

Pixel 11 also introduces Creator Suite, a comprehensive toolset designed to simplify content creation on the device. This includes social media gridlines, an on-screen teleprompter, and organizational support for captured videos. Additionally, users can now set video as their default mode when opening Pixel Camera, making it easier than ever to capture high-quality footage.

Instant Night Sight has been significantly improved in the latest iteration of Pixel 11 Pro and XL models. Users will be able to capture low-light images up to 4.5 times faster while maintaining exceptional quality. This feature is particularly useful for photographers who often find themselves shooting in challenging lighting conditions.

The new Circle to Search function allows users to instantly identify objects or species by pointing their camera at them. For example, if a user is exploring a botanical garden and points their camera at a flower, they can quickly access information about the plant’s characteristics, planting tips, or more. This feature leverages AI-generated image recognition technology.

Live Translate enables real-time speech-to-speech translation of videos and audio content using advanced on-device generative AI models powered by Google Tensor G6 processor. This innovative tool automatically dubs media from around the world into a user’s preferred language without interrupting their viewing experience.

The Pixel 11 lineup is now available for preorder, starting at $899 for the base model and reaching up to $1,299 for the Pro XL variant. The devices will be released on August 20 through Google Store and retail partners.

Nutanix has introduced the Model Context Protocol (MCP) server, a new component of its Nutanix Cloud Platform (NCP), designed to enable customers to safely use artificial intelligence tools to automate their daily cloud operations. The MCP server is open-source software that implements the Model Context Protocol specification and allows AI agents and developer tools to interact with NCP through the Nutanix Prism v4 API.

The launch of the MCP server marks a significant step in simplifying complex cloud management tasks, which traditionally require deep technical expertise and strict access controls. With this new solution, IT teams can speed up daily operations across large hybrid cloud environments without losing visibility or control.

Operating modern hybrid cloud environments is often challenging due to the need for precise infrastructure API actions that are both safe and authorized. AI assistants like GitHub Copilot, Claude Code, and Cursor can drastically speed up system management, but enterprises require certainty about the safety of AI actions. The MCP server solves these challenges by acting as a secure passthrough between AI tools and NCP.

The solution translates everyday language requests into precise infrastructure API actions while automatically enforcing existing corporate security and access policies. This allows IT teams to confidently automate routine operations and build custom AI-driven workflows, giving them more time for strategic tasks rather than manual system management.

Nutanix executive vice president Thomas Cornely emphasized the company’s commitment to using automation to simplify complex cloud operations. ‘The MCP server is an exciting next step on our journey,’ he said, ‘allowing us to bring the power of autonomous AI agents to cloud management.’

The MCP server builds directly on the Nutanix Prism V4 API Gateway, allowing connected AI tools to inherit native governance and fine-grained access controls from NCP. This ensures that enterprise AI operations are secure and predictable.

Architectural Governance and Developer Capabilities

The Nutanix Prism V4 API Gateway strictly handles all execution, governance, and security controls, enabling the MCP Server to provide a highly secure environment for enterprise AI operations. It is designed to support various developer tools and capabilities, including building automation scripts in preferred languages such as Python, Go, Java, JavaScript, PowerShell, curl, or any REST and JSON-compatible format.

For developers, the MCP server makes building automation tools faster and more intuitive by providing exact system blueprints upfront. This allows teams to instantly generate platform-ready scripts and easily configure precise security permissions through a single interface.

Customers can download the MCP server for NCP from Nutanix’s developer portal and use it with any of their own AI agents, giving them flexibility in choosing which tools best suit their needs.

Nutanix is a hybrid cloud leader offering organizations a unified infrastructure software platform to safely run AI applications and data anywhere. Trusted by customers worldwide, the company empowers over 50% of the Global 2000 to innovate faster with AI while modernizing infrastructure, simplifying operations, and controlling costs.

Artificial intelligence has become one of the biggest priorities in executive suites, with organizations investing heavily in new capabilities and employees experimenting with AI daily. Technology leaders are under pressure to identify opportunities that improve productivity and reduce costs. However, before asking how AI can speed up processes, it’s essential to question whether the workflow still makes sense.

In many organizations, the first question is ‘What can we automate?’ This might seem like a logical starting point, but I believe it’s the wrong approach. Too often, companies use AI to automate workflows designed years ago for a different business environment. These workflows have accumulated unnecessary approvals, duplicate activities, manual handoffs, and outdated policies over time.

AI may execute those processes faster, but it does nothing to address underlying complexity. This challenge is not unique; Harvard Business Review’s article ‘The secret to successful AI-driven process redesign’ highlights that organizations create the greatest value when they rethink business processes before applying AI, rather than simply layering technology onto existing ways of working.

MIT Sloan’s article ‘How AI is reshaping workflows and redefining jobs’ also emphasizes that AI delivers its biggest impact when organizations redesign how work flows across the enterprise instead of focusing only on automating individual tasks. These findings reinforce an important lesson for leaders: before asking where AI belongs, ask whether the workflow itself still makes sense.

Every business process reflects yesterday’s decisions. Most processes were never designed from beginning to end; they evolved over many years as organizations expanded into new markets, acquired businesses, introduced new systems, responded to audits or adapted to changing regulations. Each change made sense at the time but collectively creates unnecessary complexity.

For example, consider a purchasing process that requires six approvals before an order can be placed. One approval may have been added after an audit; another resulted from an acquisition; and a third was introduced because one business unit wanted additional oversight. Eventually, those approvals simply become ‘the way we do things.’ AI can summarize purchase requests, route approvals automatically, notify managers, and even recommend decisions.

What it cannot determine on its own is whether six approvals are still necessary. That requires leadership. The same pattern exists throughout finance, manufacturing, supply chain, human resources, customer service, and countless other business functions. Organizations often focus on making individual activities faster while overlooking opportunities to eliminate activities altogether.

This is where workflow redesign becomes essential. Instead of asking how AI can automate each step, leaders should ask which steps continue to create value and which exist simply because they have always been part of the process. Sometimes the greatest improvement comes from eliminating work rather than automating it.

The organizations creating the most business value from AI tend to approach the problem differently. Rather than starting with technology, they begin with the business outcome they want to achieve. This might be reducing order cycle time, improving forecast accuracy, increasing manufacturing throughput, accelerating product development, or improving customer responsiveness.

A clearly defined objective creates a much stronger foundation than simply looking for places to use AI. Once the outcome is clear, the next step is understanding the entire workflow. Many delays occur not because individual tasks are inefficient but because work passes through too many people, systems, or approval points.

Mapping the complete process often reveals unnecessary handoffs and redundant activities that can be removed before automation is introduced. Deloitte’s research on enterprise AI adoption highlights that organizations generating the greatest business value are redesigning how work is performed rather than simply automating existing tasks.

In other words, they view AI as an opportunity to change how work gets done instead of accelerating yesterday’s approach. Leaders should also distinguish between administrative work and human judgment. AI excels at gathering information, organizing data, preparing summaries, and performing repetitive tasks but struggles with decisions requiring experience, context, creativity, negotiation, or ethical judgment.

The objective is not to replace people; it’s to remove low-value administrative work so employees can spend more time applying their expertise where it matters most. Standardization is equally important when every business unit performs the same work differently, AI solutions become more difficult to implement, maintain, and scale.

Simplifying and standardizing workflows before introducing AI creates a stronger foundation for enterprise adoption while producing more consistent business results. Finally, organizations should measure business outcomes instead of technology activity. The number of AI assistants deployed or prompts submitted may indicate adoption but does not demonstrate business value.

Leaders should instead measure improvements in cycle time, quality, customer satisfaction, operating cost, revenue growth, and employee productivity – those are the outcomes executives ultimately care about. A simple framework for AI-enabled workflow redesign involves four steps: simplify, standardize, redesign, and automate.

Temporary staffing agencies are set to benefit from a significant boost in efficiency, thanks to the acquisition of Pixi by zvoove. The Munich-based software and AI solution provider has acquired the intellectual property behind Pixi, an AI-powered onboarding and support agent that automates administrative work for temporary workers.

Pixi uses WhatsApp as its primary channel, allowing employees to complete their own onboarding and manage day-to-day administrative questions directly through a chat interface. This streamlined process captures master data without requiring recruiter involvement, freeing up staff time for more high-value tasks.

The acquisition adds another powerful AI agent to zvoove’s platform, which already provides software solutions for the temporary staffing, cleaning services, and personal security industries. With Pixi on board, customers can expect a significant reduction in administrative workloads, allowing them to focus on building stronger relationships with clients and candidates.

Pixi’s capabilities are centered around natural-language processing, enabling it to understand and respond to employee queries based on customer-specific ERP data. This includes handling personal details, contracts, assignments, and working hours – all without the need for manual intervention or paper forms.

The acquisition marks a significant step forward in zvoove’s mission to digitize and optimize processes for its customers. By automating administrative work through AI-powered tools like Pixi, staffing agencies can reduce costs, increase efficiency, and improve profitability.

According to Oliver Muhr, CEO of zvoove Group, the addition of Pixi will have a direct impact on customer productivity: ‘Every process that runs without a phone call or paper form translates directly into time saved and expanded margin.’ This sentiment is echoed by Felix Huemer, CEO of zvoove Switzerland, who notes that Pixi’s automated onboarding capabilities will free up staff to focus on more human-centric tasks.

zvoove Group has established itself as the leading provider of software and AI solutions for its industry. With over 6,500 customers across Europe and Latin America, it manages a significant portion of the temporary workforce – approximately 3 million workers, with annual payroll exceeding EUR 24 billion. The company’s platform also handles over 3 million applications per year.

The acquisition is set to bring Pixi to zvoove’s customer base in Switzerland, marking an important milestone for the company and its customers. With this move, zvoove continues to push the boundaries of AI-powered automation in staffing agencies, streamlining processes and driving efficiency across the industry.

The South Korean military is taking steps to automate more of its field maintenance and logistics operations, leveraging artificial intelligence, robotics, and autonomous systems. This move comes as the army prepares for a sustained decline in available personnel due to demographic changes.

As part of this effort, Army Headquarters has commissioned a study on converting key maintenance functions to automated systems. The project is tied to ArmyTiger L, which aims to transform its future force structure through advanced technologies. Specifically, it seeks to build what the army calls a ‘smart field maintenance support system’.

The need for automation in logistics and maintenance duties is pressing due to demographic decline. As personnel numbers steadily decrease, the military must make more efficient use of its workforce. Moreover, there’s a widening gap between advances in combat platforms and the systems used to maintain them.

While newer tanks, self-propelled howitzers, and other weapons systems require fewer crew members to operate, maintenance and support operations have not undergone a corresponding reduction in staffing requirements. This discrepancy is driving the army’s push for automation.

The study will examine AI-based predictive maintenance using small language models and machine learning. Additionally, it will assess ways to automate maintenance information management. The military also plans to evaluate civilian technologies such as maintenance robots, automated tools, and wearable strength-assist suits designed to reduce physical burdens on personnel.

For logistics operations, the army is looking at automated storage and handling systems, drone delivery, autonomous transport vehicles, and palletized load systems. Technologies will be evaluated according to unit type and echelon, including infantry, armored, mechanized infantry, and artillery units.

Some of these technologies are already being tested in various capacities. The Army’s 1st Supply Group unveiled a smart logistics center last week that has automated about 70 percent of its logistics processes. This system reduced on-site workforce requirements by approximately 16 percent while cutting average processing time from eight days to five.

Trials involving wearable robotic suits have been underway at ammunition units since July, and transport drones and palletized load systems are being examined for broader military logistics applications.

A military official stated that the study aims to distinguish between facilities and equipment that can be introduced immediately and capabilities that should be pursued over the medium and long term. The army also plans to draw up short- and mid-term implementation plans by 2040, outlining what a fully developed smart field maintenance unit should look like.

Cisco has released its Stack Automation platform, which automates the deployment of GPUs, AI models, applications, and other infrastructure across hybrid environments. This new tool is designed to simplify the complex process of setting up modern infrastructure for enterprises.

The development of Stack Automation was a collaborative effort between Cisco and Quali, with the exclusive right to offer it through Cisco’s channels. The primary goal is straightforward: make setting up and managing AI infrastructure less manual and time-consuming, especially when multiple technologies are involved and different teams need to be coordinated.

Stack Automation combines Quali’s automation technology with Cisco Validated Designs and security best practices. It also integrates seamlessly with other Cisco products like Intersight and Nexus Dashboard, which will eventually become part of the Cloud Control platform for AgenticOps.

The Solutions Hub in Stack Automation is a comprehensive resource featuring technologies from both Cisco and third-party vendors. This catalog spans networking, security, observability, virtualization, AI frameworks, enterprise applications, including popular tools like Splunk, Red Hat OpenShift, NVIDIA NIMs, VMware, Qumulo, and Veeam.

For channel partners, Stack Automation presents a significant opportunity to standardize and automate their deployment processes. The validated blueprints are designed to simplify infrastructure planning and deployment, allowing them to integrate existing investments in Terraform, Ansible, Python, Shell, and Helm into governed workflows without having to start from scratch.

Customers increasingly ask solution providers to set up complex environments that span multiple areas such as networking, compute, storage, security, virtualization, and applications. By automating the deployment of these full-stack environments, Stack Automation can save businesses weeks or even months spent on infrastructure setup before they can begin using AI.

The platform’s ability to automate complex deployments while incorporating the collective expertise and validated designs of Cisco and its partner ecosystem into every workflow is a key benefit, according to Jeremy Foster, senior vice president and general manager of Cisco Compute. Customers should not have to spend weeks assembling infrastructure before putting AI to work.

Stack Automation offers different tiers for users: Essentials supports standard software deployments on a subscription basis, while the Advantage subscription adds features like Blueprint Designer, GitOps integration, advanced customization, and agentic design capabilities. Full-stack solutions, including Cisco AI PODs, are scheduled for availability in October.

The automation of complex deployments through Stack Automation has the potential to save businesses weeks or even months spent on infrastructure setup before they can use their AI tools effectively.

Paul Redding, head of MSP partnerships at NinjaOne, recently addressed a room full of managed service providers (MSPs) at The Channel Company’s XChange August event in Washington, D.C. His message was clear: the channel’s economics have shifted, and MSPs are facing a profitability paradox. Despite bringing in more revenue, rising security, compliance, tooling, and customer demands are eating into profits and making growth harder to monetize.

The complexity of environments has grown exponentially as MSPs add more technology, tools, and applications. However, they haven’t been able to pass on the increased costs to their customers. Redding noted that nobody is comfortable raising prices by 30 percent or more, which would be necessary to offset these rising expenses. This paradox is a pressing concern for MSPs looking to maintain profitability.

One of the main drivers behind this shift in economics is the proliferation of separate tools and interfaces. NinjaOne aims to eliminate this complexity by consolidating technology stacks and maximizing built-in automation. Redding emphasized that every single agent added increases risk factors, requires additional clicks, multifactor authentication, and time – all of which can be detrimental to profit margins.

Automation is a crucial aspect of addressing the profitability paradox. By automating services, MSPs can streamline their operations, reduce costs, and improve efficiency. This allows them to focus on high-value tasks rather than getting bogged down in manual processes. Redding stressed that specialization is also key – by focusing on specific vertical markets, MSPs can develop deep expertise and build stronger relationships with customers.

Gabriel Sierra, founder of Consultiply, agreed that many MSPs are looking to consolidate technology stacks and maximize built-in automation rather than layering multiple platforms onto the same customer environment. He noted that having one customer with five different tools to achieve a single goal is unproductive and inefficient. Redding’s message resonated with Sierra, who emphasized the importance of reducing the number of systems technicians must manage and making better use of existing automation within platforms.

Redding also highlighted the role of artificial intelligence (AI) as an efficiency tool rather than an immediate replacement for technicians. He sees AI as a means to improve productivity, reduce costs, and enhance customer satisfaction – ultimately leading to increased revenue growth. The ultimate goal is for NinjaOne to become a partner in MSP growth rather than just another vendor.

The difference between a vendor and a partner lies in their approach to business relationships. A vendor sells products or services without considering the long-term implications, whereas a partner takes a more holistic view – mitigating risk, improving efficiency, and growing revenue. Redding’s vision for NinjaOne is to become a trusted partner that helps MSPs make more money by streamlining operations, reducing costs, and increasing customer satisfaction.

In conclusion, the profitability paradox facing MSPs requires a comprehensive approach that addresses automation, specialization, and centralization. By consolidating technology stacks, maximizing built-in automation, and focusing on specific vertical markets, MSPs can reclaim their growth trajectory and maintain profitability in an increasingly complex environment.

A new report from Australia’s online safety regulator has revealed that nearly four in five children aged 10 to 17 have used an artificial intelligence (AI) assistant. The research, which surveyed over 1,950 children living in Australia, found that these tools are being increasingly relied upon for personal and emotional support, rather than just schoolwork.

The report, titled ‘Talking to Machines: Children’s Experiences with AI Assistants and Companions’, was conducted between February 2nd and March 4th, 2026. It highlights the growing trend of children using AI assistants as a means of seeking advice, reassurance, and conversations about personal issues.

According to the findings, around 78% of surveyed children had used an AI assistant, while 8% had used an AI companion designed specifically for emotional support. ChatGPT was found to be the most widely used service, with approximately 70% of surveyed children having utilized it at some point.

The usage of AI assistants varied by age group, with around 87% of teenagers aged 13-17 having used these tools compared to 64% of younger children (10-12 years old). This suggests that older children are more likely to be using AI assistants for a broader range of purposes beyond just schoolwork.

Most children who had used an AI assistant or companion reported doing so for functional reasons, such as accessing information, learning new skills, or translation. However, many also used these tools for entertainment and socialization, with around 68% reporting that they used them for leisure activities.

A significant proportion of surveyed children (54%) reported using AI assistants for personal and social reasons. Among these uses, a notable number asked AI for advice on physical health (33%), what to do in specific situations (33%), or their mental health and wellbeing (20%).

The report’s findings arrive at an important time as Australia’s regulatory framework for AI services affecting children has shifted from design to enforcement. Under the country’s online safety codes, certain generative AI, chatbot, and companion services must implement safeguards to prevent exposure to harmful material.

These safeguards cover both unlawful content and age-restricted material, including self-harm, disordered eating, and violence content. The age-restricted material codes came into effect on March 9th, 2026, and eSafety is now requiring providers to demonstrate compliance with these regulations.

Companies that fail to comply face significant penalties of up to $54.6 million in civil fines. In addition to this regulatory pressure, the report highlights concerns around children’s interactions with AI tools, including potentially inappropriate or harmful experiences.

One in five surveyed children who used an AI assistant or companion reported encountering a problematic interaction, such as controlling behavior, advice on self-harm, bullying, or exposure to hateful content. Furthermore, approximately one-third of these users shared personal or sensitive information with the AI tool, including photographs and secrets.

The eSafety Commissioner, Julie Inman Grant, emphasized that parents are navigating rapid changes in technology alongside other challenges associated with raising children today. She noted that AI is evolving quickly, and children are finding new ways to incorporate it into their daily lives.

Grant also pointed out that many parents recognize the benefits of using AI tools for homework or answering questions but highlighted concerns around safety by design. She called on technology providers to prioritize embedding strong safeguards into services, discouraging unnecessary sharing of sensitive information, reducing the risk of harmful interactions, and considering children’s developmental needs throughout their development.

Safety by Design has never been more important,

Researchers have found that AI mortgage assistants are prone to errors when it comes to identifying foreign deposits in bank statements. In a recent study, three leading general-use AI models were tested on their ability to flag transactions of foreign origin. The results showed that the models struggled significantly with this task, highlighting concerns about the accuracy and bias of these systems.

When researchers asked the AI models which deposits could be of foreign origin, they found that transactions connected to English names were flagged only 13.3% of the time. However, when the deposit came from a non-English name, the number jumped to 77%. This disparity suggests that the models are relying too heavily on linguistic cues rather than actual evidence.

This issue is not just about AI’s ability to provide accurate information; it also raises concerns about bias and fairness in mortgage lending. Matthew Toles, a Columbia University doctoral student and co-author of the study, notes that ‘you aren’t determined whether you’re foreign-based on how your name sounds.’ This means that borrowers from non-English speaking countries may be unfairly flagged as high-risk or denied loans.

The study’s findings are particularly concerning given the widespread adoption of AI mortgage assistants in the industry. More than 80% of lenders were evaluating the technology as of June, according to a survey by The Mortgage Collaborative, while 17% had already deployed it in live production workflows. This rapid adoption has raised concerns about the lack of standardization and testing for these systems.

Diane Yu, co-founder and CEO of Tidalwave, a mortgage technology company that collaborated with Columbia researchers on the work, notes that ‘everybody is using AI, but nobody really understands how to use AI in compliance and with the correct guardrail.’ This highlights the need for standardized testing and evaluation methods to ensure that these systems are used correctly.

The study introduces MortarBench, an open-source benchmark designed to measure the accuracy of AI mortgage assistants. The benchmark includes a set of common and useful tasks in loan origination, such as identifying payroll deposits, large transactions, and jointly held accounts. By using this benchmark, lenders can compare the performance of different models and identify areas for improvement.

The researchers tested three leading general-use AI models - Gemini 3.1 Pro, GPT-5.5, and Claude Sonnet 4.6 - on a set of tasks from MortarBench. The results showed that even these strong models struggled with accuracy, particularly when it came to identifying specific transactions in bank statements.

Zhou Yu, an associate professor at Columbia and study co-author, compares the task of finding specific transactions to ‘finding the needle in the haystack.’ However, the researchers found that the AI models often made the opposite mistake, pulling in transactions that didn’t belong. This highlights the need for more accurate and reliable methods for identifying relevant information.

The most common problem with the AI models was misclassification, where they incorrectly identified a transaction as belonging to a certain category. For example, Gemini counted a personal loan as a buy-now-pay-later transaction, while GPT-5.5 assumed all wire transfers were international. These errors demonstrate that simply finding relevant information is not enough; the systems must also accurately classify and interpret it.

Matthew Toles notes that these scores are for ‘naive use of foundational models,’ which means they are using a basic model without any additional training or fine-tuning. He cautions that major industry players typically develop their own proprietary models, which perform better on these tasks. For example, Tidalwave scored 95% on yes or no questions in a separate benchmark test.

The study’s findings have significant implications for the mortgage lending industry, particularly when it comes to AI adoption and standardization. As lenders look to automate processes and reduce costs, they must also ensure that their systems are accurate and reliable. The MortarBench study provides a much-needed tool for evaluating these systems and identifying areas for improvement.

The results of this study could have far-reaching consequences beyond the mortgage lending industry. As users increasingly rely on AI models for financial decision-making, errors and biases can carry higher stakes. Matthew Toles warns that ‘we don’t have visibility into what these models are doing, what companies are doing with them, and what the outcomes are.’ This highlights the need for greater transparency and accountability in AI development and deployment.

The study’s authors emphasize the importance of asking lenders about their use of AI mortgage assistants and whether they have undergone independent evaluation. Borrowers should also be aware that these systems may not always provide accurate or unbiased information. As Diane Yu notes, ‘you should ask those questions’ and ‘be very careful.’

The MortarBench study provides a valuable resource for evaluating the accuracy and reliability of AI mortgage assistants. By using this benchmark, lenders can compare their models to industry standards and identify areas for improvement. This could help mitigate systemic risks associated with widespread adoption of these systems.

In conclusion, the findings of this study highlight the need for greater standardization and testing in the development and deployment of AI mortgage assistants. As the industry continues to adopt these technologies, it is essential that we prioritize accuracy, reliability, and fairness.

A world of artificial intelligence-generated video has emerged, making it easier to create high-quality content quickly. However, creators face a critical question: can they use this technology commercially? The answer isn’t simple and requires understanding the terms and conditions of each tool.

Some AI tools allow commercial use on specific paid tiers, but the exact rules vary between platforms. This is where many beginners make mistakes – assuming one tool’s rules apply to all others. In reality, some grant ownership while others provide a license for its use or restrict it entirely below a certain plan.

The distinction between owning and licensing content can be complex. Ownership means that the platform doesn’t claim what you’ve generated as their own. A license allows you to use the output under specific conditions but retains rights to the underlying content or limits its use. Both types of agreements allow commercial use, but they’re not interchangeable.

To understand this better, let’s take a closer look at six tools I regularly use: Runway, Kling AI from Kuaishou, Google Veo accessed through the Gemini API or Vertex AI, ElevenLabs, Higgsfield, and Pika. Each tool has its own set of terms regarding commercial use.

Runway’s terms state that they don’t claim ownership of inputs or outputs and allow commercial use across all plans – Free, Standard, Pro, Max, and Unlimited. However, output on the free tier carries a watermark that can be removed starting with the Standard plan.

A key thing to note is that Google Veo’s terms tie indemnification to whether the model is in General Availability or if you’re using it on a paid tier. This means Preview features like Veo 3.1 may leave your project at risk of losing coverage. ElevenLabs includes a commercial license on all paid plans but requires documented consent for voice cloning and bans impersonating political figures regardless of authorization.

Higgsfield states that they don’t claim ownership or restrict commercial use across its primary terms, though this isn’t tiered by plan in the main terms. Pika licenses its service for personal, non-commercial use except where your subscription Plan expressly permits commercial use – only on Pro and Fancy tiers.

Each tool has its own rules regarding commercial use. It’s essential to check these before using any AI-generated content in a paid project. This includes checking the specific plan you generated on, not just your current subscription tier, as well as indemnification status for Preview or Beta features.

When working with cloned voices that aren’t your own, written consent is crucial and should be kept somewhere retrievable – don’t rely solely on memory. When using multiple tools in a single project, it’s essential to check each one’s terms independently because the rules can differ significantly between platforms.

In my experience, re-checking before every paid delivery is vital due to how often terms of service change in this space. I’ve built ScreenWeaver as a workspace between writing and production for exactly this reason – it keeps track of tool, plan tier, and rights status attached to each shot next to the script line.

The free plan does not grant commercial use even if you remove the watermark yourself. Removing a watermark is just a visual signal; the underlying license remains unchanged. If the free tier excludes commercial use, editing out the watermark doesn’t give rights that weren’t given in the first place.

Combining outputs from different tools can add complexity. Each output is still governed by its respective tool’s terms and conditions – there isn’t a single ‘AI video license’ covering mixed pipelines.

When working with AI-generated video commercially, creators need to carefully consider each tool’s terms and conditions. By following this checklist before shipping content, creators can avoid costly mistakes and ensure they meet the necessary requirements for commercial use.

Commercial use of AI-generated video is allowed in most platforms but requires a careful examination of each tool’s specific rules and conditions. Understanding these nuances will save creators from costly mistakes and ensure they meet the necessary requirements.

The process of using multiple tools to generate content can be tricky, as you need to satisfy both sets of terms individually. This means checking each platform’s rules independently will help avoid potential issues down the line.

The price tag is becoming the most important benchmark for large language models (LLMs), a trend that’s gaining momentum in the industry. This shift is evident in recent announcements from major players, including Google and Meta, which are now prioritizing cost-effectiveness over other metrics. The reason behind this change lies in the growing demand for AI tools that can be integrated into business workflows without breaking the bank.

Google has taken a significant step in this direction with its latest release of Gemini 3.7 Flash, touted as its most intelligent workforce model to date. This model is designed for coding, knowledge work, and agents, but what’s notable about it is the price point: $0.75 per million input tokens and $3.75 per million output tokens. While other benchmarks are still being mentioned, the focus on cost has become a major selling point.

This trend of prioritizing price over other metrics isn’t unique to Google; multiple model providers have followed suit in recent releases. Writer, for instance, highlighted its Palmyra X6 flagship model and upgrades to its Writer Agent harness, emphasizing the reduced costs within the harness. The company claimed that Writer Agent operates at a 52% lower cost compared to previous versions.

Meta’s move back into open weight models with Muse Glimmer has also been accompanied by aggressive pricing strategies. This is part of Meta’s comeback plan, which aims to make its AI tools more accessible and affordable for businesses. The company’s decision to prioritize price reflects the growing concern among enterprises about token costs and their impact on corporate budgets.

Nvidia has also entered this space with its Nemotron model family, designed specifically for customization by enterprises and SaaS vendors. SpaceXAI is another player that’s been aggressively pricing its LLMs, including Grok 4.6. DeepSeek has taken a different approach with dynamic pricing, which allows it to adjust prices based on demand while still undercutting other models.

The shift towards prioritizing price as the key benchmark for LLMs isn’t limited to these companies; many others are following suit. Databricks’ recent annual revenue run rate announcement highlighted the growing adoption of open-source AI by enterprises, which is driving this trend. As Ali Ghodsi, CEO of Databricks, noted in a statement: ‘Enterprises don’t just want AI that talks; they want agents working across their business that remember context, deliver accurate answers, and execute work without blowing through their budgets.’

This emphasis on cost-effectiveness has significant implications for the industry. Companies like Uber are already adjusting their token budget to reflect this new reality. Vendors are facing a ‘token budget hangover,’ as Palantir CEO Alex Karp put it, where AI tools that were once seen as revolutionary are now being scrutinized for their costs and effectiveness.

The impact of this shift is evident in the words of various CEOs during earnings calls. Moody’s CFO Noemie Heuland emphasized the importance of monitoring token costs closely: ‘Our internal AI and token cost today is actively governed… We have a variety of tools that we put at the disposals of our engineers, our back-office teams.’ Similarly, Synchrony Financial CFO Brian Wenzel noted that companies need to rethink their approach to AI deployment: ‘You may say,

The field of mathematics is undergoing significant changes as artificial intelligence (AI) increasingly tackles complex problems once thought beyond its reach. This shift has mathematicians reevaluating their role in the discipline and considering how it will evolve alongside rapidly advancing technology.

At the recent International Congress of Mathematicians in Philadelphia, 2006 Fields Medalist Terence Tao reflected on this pivotal moment for mathematics during a talk titled ‘Mathematics in the Age of AI.’ While many conversations focus on the capabilities of future AI systems, Tao argued that mathematicians need to ask more fundamental questions about their goals and values.

Tao assumed that soon, AI will be able to perform a reasonable fraction of mathematical tasks successfully. He then posed a crucial question: If this is the case, what does it mean for our understanding of successful mathematics? In other words, why do we do mathematics in the first place?

Mathematicians have long focused on problem-solving as a core aspect of their work. However, with AI now impacting this area significantly, Tao chose to focus on this facet during his talk. He identified several stages involved in mathematical problem-solving: providing and verifying proofs, evaluating and explaining these proofs, and making them accessible to the broader community.

AI has made significant strides in generating candidate proofs or crucial ideas, as well as rigorously checking very long or intricate proofs. However, Tao emphasized that AI’s limitations become apparent when it comes to human evaluation and explanation of these proofs. He noted that even large models can produce 100,000-line proofs that are difficult for humans to understand.

The problem lies in the fact that if no one understands a proof, its likelihood of being accepted by the community becomes slim. A successful result is not just about generating a correct answer but also about making it valuable and useful to others. AI might be good at finding results, but it often fails to provide context that makes people care.

Tao argued that without human evaluation and explanation, a proof cannot reach its full potential – namely, becoming canonicalized or widely accepted as part of the mathematical canon. This process involves not just generating correct answers but also providing context, describing how one arrived at those results, and making them accessible to others.

The value of mathematics is often measured by its applications in other fields such as engineering and physics. However, these applications are typically built on top of canonicalized knowledge – the kind found in textbooks rather than recent research papers. Tao noted that AI itself relies heavily on this existing body of work for its success.

So where do we go from here? Tao encourages mathematicians to engage in discussions about their goals and values regarding problem-solving, education, publication, and community culture. He challenges them to take the initiative in integrating AI into mathematics rather than passively reacting to it.

The integration of AI will require mathematicians to set clear rules on what is acceptable or not, rather than letting external actors define these standards for us. This involves considering how we can use AI tools effectively while maintaining our values and goals as a community.

A $7 billion technology company based in Silicon Valley is aggressively promoting its AI-powered tools for automating administrative tasks in healthcare, with some customers reporting significant financial losses and negative outcomes. According to internal documents and communications obtained by STAT, Commure offers substantial compensation to medical clinics and other parties willing to refer its products to new business prospects. This referral program is part of the company’s strategy to rapidly sell its AI tools into diverse clinical settings with customizable financial incentives.

Commure was founded in 2020 on a vision to build a new operating system for healthcare, leveraging artificial intelligence to take power and money from massive insurers and give it back to doctors. The company’s executives have framed their mission as a kind of Robin Hood story: using AI to redistribute wealth from industry giants to medical professionals. However, Commure isn’t just promising to enrich doctors with money from insurers; the company also offers compensation to those who promote its products.

STAT has found that Commure offers thousands of dollars in various forms to customers and other parties willing to refer its products to new business prospects. This is part of a broader effort by the company to swiftly sell AI tools it claims can unburden clinicians burned out by healthcare’s bureaucratic morass. The compensation is seen as crucial for the company’s rapid growth, with Commure claiming that its tools are used by more than 500 health care organizations across the country.

Many of Commure’s customers speak highly of the company and its products, testifying in marketing videos and podcast interviews that they have restored joy and reliable revenue to their practices. The company claims that its AI-powered tools for billing, scheduling, and clinical documentation are used by ‘130 of the nation’s largest health systems,’ including giants like HCA Healthcare and Tenet Healthcare.

However, some customers report suffering steep financial losses and other negative outcomes after implementing Commure’s products. Despite these concerns, the company maintains that its referral programs are industry standard and that the vast majority of its hundreds of customers are happy with its services. The effort to rapidly sell AI into diverse clinical settings with customizable financial incentives magnifies the difficulty of evaluating products that directly impact the cost and quality of health services received by millions of Americans.

Commure’s CEO, Tanay Tandon, has expressed his vision for a world where doctors are millionaires while industry giants like UnitedHealth have significantly reduced market value. However, this goal remains elusive as Commure continues to face challenges in balancing its business model with the needs and concerns of its customers.

The Defense Intelligence Agency (DIA) is working towards a future where artificial intelligence tools support military operations through seamless interactions between digital agents. According to Maj. Gen. Robert Kinney, the DIA’s chief AI officer, this vision involves ‘agent-to-agents’ interactions that enable data analysis and decision-making across combatant commands.

Kinney shared his insights during a panel at the DODIIS Worldwide conference in Tampa, Florida, where he discussed the agency’s ongoing efforts to build an AI enterprise platform service. This 90-day ‘sprint’ aims to establish a comprehensive framework for integrating AI tools and data sources within the military intelligence community.

One key component of this effort is the model context protocol (MCP), which will provide a standardized way to access intel data from various external systems and data sources. MCP serves as the technical foundation that enables AI tools to connect with these external systems, facilitating more efficient data exchange and analysis.

Looking ahead to the next year, Kinney emphasized the importance of building digital agents that can interact with each other in a secure and compliant manner. This involves developing ‘agents-with-agents’ capabilities, which will allow for more sophisticated decision-making and task execution within military operations.

The potential benefits of this approach are significant, as it could enable real-time data sharing and analysis between different directorates and commands. For instance, an agent responsible for collection management might communicate with agents in the 3 Directorate (operations) or the 4 Directorate (contested logistics), facilitating more informed decision-making across the board.

However, officials must also consider the risks associated with these emerging technologies. Kinney noted that setting up guardrails to mitigate potential risks is crucial, particularly when it comes to irreversible mission areas such as fires and planning. In such cases, human oversight will be essential to ensure responsible use of AI capabilities.

The DIA is currently laying the digital foundation for this vision by establishing ‘pipes’ that support data access and agent development. Additionally, officials are exploring the ‘tradecraft piece’ of using agents interacting with each other, focusing on responsible parameters for operating within these systems.

Suno has released a major upgrade to its browser-based generative audio workstation, Suno Studio 2.0. This update marks a significant shift in the company’s approach to music production, as it integrates AI-powered tools into every aspect of the workflow. The new version is available now to Premier subscribers and brings with it several key features that blur the lines between traditional digital audio workstations (DAWs) and generative audio software.

The centerpiece of Studio 2.0 is its support for MIDI recording and editing. This feature has been highly requested by users since the workstation’s launch in September 2025, and Suno product manager Henry Phipps notes that it was a crucial addition to the platform. With this update, users can import, record, and edit MIDI directly on the timeline, play notes with a typing keyboard if they don’t own a controller, and design sounds with a new built-in wavetable synth.

The integration of MIDI into Studio 2.0 is not just about recording and editing; it’s also about using that data to generate new audio. Clips can serve as prompts for AI-generated music, allowing producers to play in a chord progression or melody and have Suno render it as audio, extend it, or build on it. This approach marks a significant departure from traditional DAWs, which often rely on manual input and editing.

Suno’s chat bar is another key feature of Studio 2.0. In beta at launch, this conversational assistant ties into the live session and understands music production. Users can ask it to create instruments and vocals from scratch, generate lyrics or parts in context, or tidy up a session. The same interface also allows users to build custom audio plugins by describing them in chat.

The new plugin creation mechanism is notable because Suno has stated that it will not require credits at launch, although a credit structure may be introduced later. This approach suggests that the company wants to make its tools accessible to as many users as possible while still generating revenue from premium features and services.

Automation also makes its debut in Studio 2.0, allowing users to draw curves that alter track and plugin parameters across an arrangement. An updated stem-separation tool extracts component parts of a song at high audio fidelity, including external audio brought into Suno, which can then be split, arranged, and rearranged inside the platform.

Premier subscribers will have access to export 32-bit/48kHz multitracks and stems without limitations, while non-subscribers will still be able to use many of Studio’s features. However, the DAW-style workflow is reserved for Suno’s highest-paying subscribers, which may limit its appeal to some users.

Suno has also announced a global partnership with BMG, one of the largest music companies, tied to what it says will be its first music model developed with the industry. This deal promises new economic opportunities for artists and songwriters who opt in, as well as deeper collaboration on Suno’s programs for independent artists.

The sequencing matters because Suno has spent the past two years as the music industry’s primary litigation target. The major labels sued the company over its training data in 2024, and the fight has been ongoing since; Unite.AI covered Suno’s post-lawsuit principles reset earlier this month, which included commitments to no artist names in training metadata, no prompts for specific artists or copyrighted songs, third-party screening of uploads, and coming audio watermarking and fingerprinting.

Suno wants to be infrastructure the industry uses, not a generator the industry sues. The company’s near-term roadmap is now visible on two tracks: the opt-in, industry-developed music model promised under the BMG partnership, and the continued conversion of Studio from a generator with editing bolted on into the production environment those licensed models would run inside.

The feature list in Studio 2.0 overlaps meaningfully with entry-level DAWs, including multitrack timeline editing, MIDI, automation, effects, and export. However, it does not support third-party plugins – the synthesizer and effects are Suno’s own, whether factory-built or chat-generated – which keeps Studio a closed environment rather than a host for existing plugin ecosystems.

Suno has separately flagged changes to its downloads policy taking effect September 3, aimed at limiting mass distribution of generated songs to streaming platforms. This constraint sits alongside the export freedoms granted paying producers in Studio 2.0.

A growing number of Americans are hesitant to grant access to their personal data, hindering the adoption of AI assistants in smart homes. According to a recent report by Reviews.org, 63-65% of consumers express concern about major AI assistant platforms like Alexa and Siri. This is not just a minor issue; it’s a fundamental permission gap that tech companies are struggling to bridge.

The problem lies not with the technology itself but with how users perceive its capabilities and their own control over data collection. When 78% of users say they would disconnect a device if it collected more data than expected, and 50% have already deleted or limited their conversation history, it’s clear that trust is a critical issue.

The Wharton Blueprint for AI Agent Adoption identifies three primary frictions: perceived competence, trust, and the delegation of control. Notably, control concerns account for 26% of the total weight in the adoption decision. As Professor Stefano Puntoni notes, ‘What is difficult is the series of uncomfortable decisions that the end user and organizations need to take.’

This friction extends beyond personal use cases to the workplace as well. The IBM 2026 CEO Study highlights a massive gap between AI access and regular usage. While 85% of employees have access to AI tools, only 25% use them regularly. CEOs may believe their workforce is ready for AI assistants, but adoption depends on people, not just technology.

Consumer expectations are also at odds with the current state of AI development. The Prophet 2026 AI-Powered Consumer Report shows that 67% of consumers want AI to anticipate their needs without being asked. However, there has been a decline in the belief that consumers will rely on generative AI for most decisions. People crave convenience but are increasingly unwilling to grant permissions necessary for accurate anticipation.

The industry is attempting to bridge this gap through deeper integration and more user-centric design. For instance, Samsung’s pivot to clinically-integrated agentic health tracking systems connected to over 500 hospitals demonstrates a willingness to adapt to changing consumer expectations. However, the high stakes remain: either companies earn user trust or AI assistants will remain novelties rather than utilities.

The agent business model relies heavily on user trust. If companies prioritize aggressive data harvesting over user-defined boundaries, they risk creating sophisticated yet permanently disabled tools.

Google has announced that its artificial intelligence (AI) assistant app, Gemini, has reached a significant milestone with over one billion monthly active users. This marks the 14th Google product to hit this mark, according to CEO Sundar Pichai’s announcement via X.

The rapid growth of Gemini is notable, especially when compared to other AI-powered chatbots like OpenAI’s ChatGPT, which also reached one billion monthly active users in June. However, it’s worth noting that today’s figure specifically refers to the standalone Gemini app and does not include users from other channels or Google’s AI Mode in Search.

Google has been steadily integrating Gemini across its various products, including Search, Workspace, Android, and its standalone app. The company’s efforts have paid off, with over 1 billion monthly active users globally using AI Mode in Search alone. However, the new milestone specifically highlights the growth of the Gemini app itself.

The company continues to roll out new models and features for Gemini, including the latest version, Gemini 3.5 Flash. This update is designed to improve coding and autonomous AI-agent tasks, further enhancing the capabilities of the chatbot.

According to Google’s data, a significant portion of users are taking advantage of the voice feature when interacting with Gemini, with 63% using this method directly. Additionally, the app generates over 150 million images every day, making it a key player in AI-generated content.

The growth of Gemini is not limited to Google’s own products; the company has also seen significant adoption on iOS, with more than 100 million active users. This milestone comes shortly after Google’s Q2 2026 earnings call, where the company celebrated having over 950 million monthly users and saw daily active users triple in the past year.

The announcement precedes Google’s upcoming Made by Google event, which is expected to showcase new features powered by Gemini across Pixel devices.

Claude users are facing an unexpected change in how their work is handled by the Anthropic model. Starting immediately, all media created with Claude will include invisible watermarks that identify it as AI-generated content. This move has sparked a heated debate among users who claim they cannot opt out of this new policy.

The decision to implement these watermarks comes after Anthropic announced its compliance with European Union regulations regarding transparency in AI-generated content. The ‘Code of Practice on Transparency of AI-Generated Content’ sets rules for marking and detecting AI-created media, including text, images, audio, and video. As part of this agreement, nearly 200 companies have signed up to comply, including Anthropic.

Claude users are particularly upset because they claim the watermarks will degrade their work or make it unusable in certain situations. For instance, if a user copies and pastes text generated by Claude into another document, the watermark may not be visible due to heavy editing. Similarly, metadata attached to images created with Claude could disappear if the file format changes.

The backlash against this policy has been swift on social media platforms like X and Reddit. Users have expressed their discontent, calling it ‘hugely problematic’ and an end to ‘the era of unprovable AI writing.’ Some users who rely on Claude for proofreading text they’ve written themselves are concerned that the watermark will reveal their use of AI tools.

A radio show host and blogger, Erick Erickson, posted a scathing review on X: ‘I had ditched Grammarly for Claude because it does a better job. But now my work will be watermarked as being generated by Claude.’ His comment highlights one of the main concerns users have – that their work is no longer entirely theirs.

The debate surrounding AI-generated content has also sparked discussions about who should receive credit for this type of work. Some argue that human users should get credit, not the machine itself. Others claim that acknowledging the AI’s role in creating the content is essential and fair. One Reddit user started a vigorous discussion by stating: ‘I gave the instructions, context, decisions, and countless refinements; Claude was just the tool.’

Anthropic has emphasized that its decision to implement watermarks aligns with EU regulations and aims to increase transparency around AI-generated media. However, users are still concerned about how this will affect their work and whether it’s a necessary step towards accountability.

The use of invisible watermarks is not new; Google has been watermarking its AI-generated images since 2023 and has expanded the practice to text, audio, and video. OpenAI also uses similar technology for image watermarking but has chosen not to release it due to concerns about false positives and circumvention.

It’s worth noting that this policy change affects all Claude models and products worldwide, including Claude Code and Claude Cowork. Users cannot opt out of the new policy, which will apply uniformly across Anthropic’s offerings.

Claude, the artificial intelligence model developed by Anthropic, will soon start watermarking its generated content. This move is in compliance with EU rules, specifically Article 50(2) of the EU AI Act, which requires transparency in AI-generated content. The company has signed the Code of Practice on Transparency of AI-Generated Content and outlined how it plans to implement these regulations.

In a support document, Anthropic explained that models launched in the European Union (EU) after August 2, 2026, will have watermarking capabilities at launch. This includes output from various Claude platforms such as the API, Code, Cowork, and Tag. The watermarks will be embedded in generated text files and include digitally signed provenance metadata for supported file types.

The watermarking feature is not limited to Europe; it will be applied globally wherever Claude is offered. However, Anthropic notes that some platforms or features might not support certain types of watermarks. This means users may still encounter AI-generated content without visible watermarks if they don’t know what to look for.

For text files, the watermark will be integrated into the generated text itself and travel with it when copied or pasted elsewhere. For images and other file formats, Anthropic plans to use the Coalition for Content Provenance and Authenticity (C2PA) open standard. Details on recognizing these watermarks are expected in forthcoming documentation.

Existing models launched before August 2, 2026, will also receive watermarking support at a later time. However, it’s worth noting that watermarking may not be foolproof. Heavily edited or translated text might not carry detectable watermarks, and metadata can be stripped through format conversion or other means.

Anthropic’s move follows similar commitments by AI companies like OpenAI, Google, and Meta. This development aims to increase transparency in AI-generated content and help users identify the origin of such material.