Latest News

A new era in cybersecurity has dawned, one where autonomous systems can reason, adapt, and operate continuously. However, this shift also brings significant challenges, as attackers can now generate exploits faster, scale campaigns further, and operate with unprecedented efficiency. The traditional approaches to security are no longer sufficient, and a new Cyber Stack is needed to keep pace with the evolving threat landscape.

The physics of cybersecurity have changed dramatically in recent years. With the rise of AI-powered attacks, defenders must adapt their strategies to stay ahead of the curve. Microsoft has been working on Project Perception, a comprehensive agentic security system designed specifically for this new reality. At its core, Project Perception is built around three key principles: continuous perception, reasoning, and action.

Project Perception brings together signals, context, models, and specialized agents into a continuously learning system of defense. This integrated approach enables the Cyber Stack to reason across vast amounts of context, prioritize actions at machine speed, and take corrective measures in real-time. The system is designed to amplify defenders with better insights and more powerful ways to act.

The defining characteristic of Project Perception is its ability to continuously perceive risk across the entire digital estate. This involves a deep understanding of how an attacker sees the world, how a defender evaluates risk, and how protections are improved over time. To achieve this, Project Perception coordinates three classes of specialized agents: Red team agents identify potential paths to compromise before an attacker can exploit them.

Blue team agents investigate, reason over context, and determine what represents meaningful risk. Green team agents take corrective actions and strengthen defenses across the environment. Working together, these agents form a closed-loop system that continuously discovers, evaluates, and improves an organization’s security posture. This integrated approach enables Project Perception to stay ahead of emerging threats.

A key aspect of Project Perception is its ability to provide broad visibility across the digital estate. Microsoft brings together signals from various sources, including identities, endpoints, applications, data, clouds, and AI systems. This comprehensive view allows agents to reason over risk in real-time, prioritize actions, and make decisions that are informed by a deep understanding of the environment.

Project Perception’s multi-model architecture is another critical component of its design. Rather than relying on a single model, Project Perception combines frontier and specialized cyber models, optimizing for both quality and cost. This approach enables customers to benefit from advances in AI without being tied to any single model. The system continuously selects the capabilities best suited to the task, ensuring that each security workflow is optimized for effectiveness and economics.

The actuators within Project Perception are designed to connect insights to actions. Security teams do not need more information; they need better outcomes. Actuators enable agents to take corrective measures in real-time, reducing risk rather than simply identifying it. This integrated approach empowers defenders to strengthen security while remaining firmly in control.

Project Perception is built with safety first principles at its core. The system inherits the security, compliance, governance, and operational controls that Microsoft’s customers already rely on. This ensures that capabilities are delivered with the same rigor, accountability, and enterprise readiness that customers expect.

The future of security will be shaped by AI-powered attacks. Defenders need systems that can continuously perceive, reason, and act alongside them. Project Perception is a significant step towards building this future. By providing a comprehensive agentic security system, Microsoft aims to help organizations stay ahead of emerging threats and protect their digital estates with confidence.

Project Perception enters public preview on August 3, marking an important milestone in the development of this new Cyber Stack. As the threat landscape continues to evolve, it is clear that traditional approaches to security will no longer suffice. Project Perception offers a comprehensive solution for defenders seeking to stay ahead of AI-powered attacks and protect their organizations with confidence.

Microsoft’s commitment to building a secure future extends beyond Project Perception. The company has been working on various initiatives aimed at enhancing AI security through global red teaming, research-to-reality projects, and other efforts. These endeavors demonstrate Microsoft’s dedication to staying at the forefront of cybersecurity innovation and its willingness to collaborate with experts in the field.

Project Perception is a testament to Microsoft’s ongoing investment in AI-powered security solutions. By leveraging machine learning jobs and expertise from across the industry, Project Perception represents a significant step towards building more effective defenses against emerging threats.

ChatGPT slash commands have been making the rounds on social media, with many claiming to unlock hidden features and improve answer quality. But what’s real and what’s just hype? Let’s take a closer look at these so-called ‘secret’ commands and see if they live up to their promises.

When you come across prompts like /handwritten or /humanize, it’s easy to assume that they’re official interface shortcuts. However, most of these viral labels are simply ordinary words preceded by a slash – not actual software commands. The distinction matters because ChatGPT does have some genuine interface shortcuts, but a viral prompt doesn’t become an official feature just because it starts with /.

To tell the difference between real and fake commands, let’s start with what we know about ChatGPT’s interface. A genuine slash command is connected to the ChatGPT interface and normally appears in the suggestion menu when you type /. Availability can vary by platform, account, and rollout, so the menu visible in your own composer is the safest current list.

OpenAI officially documents /canvas for opening Canvas, although its help page describes it as a backslash command in some places. OpenAI also documents /study where slash suggestions are available, while the more broadly supported shortcut is @study. Older macOS release notes mention slash shortcuts for actions including search, reasoning, and image generation – but these are interface actions, not secret instructions that improve every answer.

A prompt formula that works better than a secret command is to be clear, specific, and sufficiently detailed, then refining the request after reviewing the result. A useful reusable formula is: Task + context + audience + constraints + output format + verification. For example, ‘Compare these two phones’ is workable, but the following is much more controlled: Compare Phone A and Phone B for an Indian buyer with a ₹30,000 budget. Use current official specifications and India prices. Include a compact table, explain the display, performance, cameras, battery, and update policy, flag any unverified claim, and finish with separate recommendations for gaming and photography.

Eight useful prompt templates to save are: 1) Natural rewrite without changing facts – ‘Rewrite the text for a general reader in a warm, direct tone. Preserve names, numbers, quotations, and factual meaning. Vary sentence length, remove repetition, and list every material change you made.’; 2) Simple explanation – ‘Explain this to a reader who knows the basics but not the jargon. Use one everyday analogy, define technical terms on first use, and finish with a three-point recap’; 3) Reliable summary – ‘Summarise this in five bullets of no more than 20 words each. Preserve dates, prices, and warnings. Add a separate ‘What this does not say’ line’; 4) Product comparison – ‘Compare these products using only the supplied data and linked official pages. Create a table, name a winner for each category, separate claimed figures from measured results, and explain who should buy each one’; 5) Fact-check a viral claim – ‘Check this claim against current primary sources. Separate confirmed facts, missing context, and unsupported statements. Provide source links and publication dates. Say clearly when evidence is insufficient’; 6) Improve an email – ‘Rewrite this email to be concise and polite without sounding stiff. Keep the request and deadline explicit. Do not invent background information. Give me a subject line and the final email’; 7) Turn notes into an action plan – ‘Convert these notes into tasks. For each task include the owner, next action, deadline, and dependency. Put unresolved decisions in a separate questions section’; 8) Research before writing – ‘Research this topic using current primary sources first. Show me the proposed angle and key facts with links before drafting. Avoid copying source wording and identify any claim that could not be independently verified’.

When it comes to testing a command, here’s what you can do: type the slash and wait – if ChatGPT displays a selectable interface suggestion, it is likely a supported shortcut for your account; check official documentation – social posts can remain online long after features change; try it without the slash – if ‘humanize this’ produces the same behavior as /humanize, you are seeing normal language interpretation rather than a special mode; and inspect the result – a confident answer can still contain invented facts, citations, or capabilities.

For recurring preferences such as tone, preferred units, or response structure, ChatGPT’s Custom Instructions are more suitable than repeatedly pasting a made-up slash command. For high-stakes medical, legal, or financial decisions, always verify the answer with qualified sources.

There is no universal catalogue of secret ChatGPT slash commands – and that includes /handwritten and /humanize. A few slash shortcuts can open real tools like Canvas or Study Mode on supported interfaces, but these viral labels are generally just compact prompts. They may appear to work because ChatGPT understands ordinary language – not because they unlock a hidden feature.

The smarter shortcut is a reusable prompt that clearly states the job, audience, constraints, format, and verification standard. It takes a few more words, but it gives ChatGPT far less room to guess.

Cornerstone University has launched the President’s Artificial Intelligence Advisory Board, a new initiative aimed at guiding the university in its integration of artificial intelligence across all aspects of student experience and operations. The board brings together distinguished leaders from technology, industry, and higher education to provide strategic counsel, technical insight, and ethical grounding for Cornerstone’s AI endeavors.

The launch was announced on Saturday through an exclusive with FOX News, highlighting the university’s commitment to being at the forefront of AI adoption in Christian higher education. According to Dr. Gerson Moreno-Riaño, president of Cornerstone University, ‘Artificial intelligence is the most revolutionary technology in human history.’ He emphasized that it has already impacted every aspect of life and is reshaping fundamental aspects of the human experience.

Cornerstone’s President’s AI Advisory Board will focus on providing a comprehensive framework for understanding and utilizing AI within a Christian context. The goal is to prepare students to harness AI for the flourishing of all, as stated by Dr. Moreno-Riaño. This initiative reflects Cornerstone’s mission to educate tomorrow’s Christian influencers – graduates who are market-ready, morally grounded, and equipped to lead in industry and ministry.

The board brings together a diverse group of experts with expertise spanning artificial intelligence, industry innovation, ethics, workforce development, and emerging technologies. Founding members include Dr. Robert J. Marks, Senior Fellow/Director of the Bradley Center for Natural & Artificial Intelligence at Discovery Institute and Distinguished Professor of Electrical and Computer Engineering at Baylor University.

Dr. Don Barger, Director of Innovation and Artificial Intelligence for the International Mission Board, is also part of the board. Additionally, David Copps, CEO and Co-founder of Worlds, Dr. Angus J. L. Menuge, Professor and Chair of Philosophy at Concordia University Wisconsin, Dr. William A. Dembski, Founding Senior Fellow of Discovery Institute’s Center for Science & Culture and Distinguished Scholar for Mathematics and Philosophy at Cornerstone University, Dr. Onsi Fakhouri, Software Architect, Technology Executive, and Scientific Researcher with Haiku Creations, and Jonathan Swindell, Ph.D. Candidate, Baylor University Department of Electrical and Computer Engineering are all members.

As AI becomes increasingly integrated across industries, employers will seek individuals who possess expertise in applying technology to solve problems and create opportunities. Cornerstone is preparing students for this need by emphasizing the importance of a Christian worldview framework when utilizing AI. The university’s Applied Artificial Intelligence Certificate program is designed to help students develop AI fluency and workplace-ready skills.

The certificate program focuses on practical, hands-on learning experiences that emphasize how AI is applied across various fields such as communication, leadership, research, decision-making, and professional environments. Unlike many other AI programs, Cornerstone’s emphasis is on the Christian worldview framework for understanding and using AI. This approach aims to equip students with both technical skills and moral grounding.

Dr. Robert Marks will serve as co-chair of the President’s AI Advisory Board alongside Dr. Gerson Moreno-Riaño. As a senior fellow and director of the Bradley Center, Marks has extensive experience in AI research and development. He is also a fellow of several prestigious organizations including the Institute of Electrical and Electronic Engineers, Optica, and the Asia-Pacific Artificial Intelligence Association.

Dr. Marks expressed his excitement about being part of Cornerstone’s initiative to integrate AI within a Christian context. ‘An education that ignores artificial intelligence leaves students ill-equipped for the modern world,’ he stated. He emphasized that an education must address both technical skills and moral grounding when it comes to AI adoption, particularly in a Christian higher education setting.

The President’s AI Intelligence Advisory board and AI certificate reflect Cornerstone University’s commitment to being The Destination of Choice for Christian Higher Education. This initiative demonstrates the university’s dedication to preparing graduates who are equipped to lead in industry and ministry while navigating one of the most significant technological shifts in human history.

A simple Google search turned up a big problem for Claude AI - conversations from the popular chatbot started showing up in search results. Users on Reddit stumbled upon this issue and sounded the alarm quickly.

Some of these chats contained sensitive information, like financial or legal details. But that was just the tip of the iceberg: a more pressing concern was how easily users could access these conversations using site:claude.ai/share - everything from fanfiction to news discussions and programming topics was exposed online.

When Claude AI’s creators at Anthropic were alerted about the issue, they moved fast. They shut down links that revealed sensitive information and removed related search results from Google. But many of those chat links are still live online, leaving users’ private conversations vulnerable.

This isn’t an isolated incident - other AI chat platforms have had similar problems over the past year. The issue here is a lack of proper indexing controls for shared conversations on Claude AI. This raises serious questions about the security of these online interactions and whether they’re as secure as people think they are.

Claude AI does offer users a way to share chats, but there was no safeguard in place to prevent search engines from accessing them. That oversight put potentially sensitive information at risk - not exactly what you want when using an AI assistant like Claude for private conversations.

If you’re worried about your chat security, head over to Settings > Privacy > Shared Chats and tap Manage. Check if any of your shared conversations are exposed online. If so, consider disabling active sharing links until a more thorough solution is implemented - it’s the least we can do to keep our conversations safe.

A temporary fix won’t cut it - Anthropic needs to tackle the root cause of this issue rather than just treating its symptoms. They should implement proper controls and disable shareable links server-side to prevent similar problems in the future, or at least until they’re sure everything is secure.

The company’s response will be key in rebuilding trust among users who rely heavily on chatbots like Claude for sensitive interactions - it won’t happen overnight, but with a clear plan of action, users can start feeling safer again. This incident is a stark reminder that security and privacy must always be top priorities, even with AI-powered tools.

Claude AI users have shown great vigilance in uncovering this problem - their findings highlight the importance of community involvement when it comes to online security. The Reddit community has taken an active role in making sure online interactions remain secure, which is something we should praise and build on, not just criticize for pointing out issues.

This incident serves as a warning about the need for robust security measures on AI platforms like Claude - users must be vigilant when interacting with these tools to ensure their safety. The conversation started by Reddit users will continue in earnest now that they’ve raised awareness of this issue and its potential consequences.

Anthropic should provide clear answers about how they plan to address this issue beyond just temporary fixes. Their response will be key in rebuilding trust among users who rely on chatbots for sensitive conversations, and we need concrete steps before anything else can move forward - vague promises aren’t enough anymore.

The security of AI-powered tools like Claude is a shared responsibility between the creators and their users - both must work together to ensure online interactions remain secure. By acknowledging this and taking concrete action now, Anthropic can start rebuilding trust with its users.

A recent issue has come to light where private conversations between users and Anthropic’s AI chatbot, Claude, were inadvertently exposed in public Google search results. The affected chats included sensitive information such as cryptocurrency wallet keys and personal details like names and addresses. This incident highlights the potential risks of sharing sensitive data with AI-powered tools, particularly when it comes to data analysis tools for businesses that rely on these technologies.

The rapid advancements in artificial intelligence (AI) have been a topic of discussion among experts and the general public. Every few weeks, we hear about new capabilities from the AI community that are pushing the boundaries of what is possible. While this has sparked excitement and curiosity, it also raises concerns about the potential impact on society and jobs.

The simultaneous government announcements of new regulatory actions have received less attention but are fundamentally important. There are over 1,500 bills under consideration in various states, hundreds in Congress, dozens of executive actions from the executive branch, and a diverse set of views among market participants on how to regulate the AI industry.

Initially, I thought that the AI revolution could be governed by the invisible hand of the market. However, with nearly 2,000 proposals being considered, it’s clear that this is not possible. While many proposed policy changes are thoughtful and necessary, they collectively fall short in addressing a crucial issue: establishing a durable, comprehensive future-focused regulatory framework.

National regulatory bodies have traditionally been established after a crisis. The Securities and Exchange Commission (SEC) was created following the 1929 stock market crash, while the Nuclear Regulatory Commission (NRC) was formed after the partial meltdown at Three Mile Island. It’s essential to learn from these examples and not wait for a crisis to occur during this AI revolution.

The establishment of a national regulatory body is not a panacea. A dedicated commission focused on AI impacts would have a dynamic mandate, as it affects various aspects of society. Regulators tend to over-regulate, so continued congressional, executive, judicial, and public oversight is critical. The AI revolution is a global competition, requiring balance between innovation-driven markets and proper regulation.

One aspect of the SEC’s operating model could serve as a partial way forward. In my previous role at Nasdaq, I was surprised by the rules that governed implementing improvements to core exchange technology. We had to submit detailed changes for public comment and subsequent review, which would then be published. This process ensured transparency but also allowed competitors to know exactly what we were planning.

This operating method has contributed to U.S. capital markets being among the best in the world. The SpaceX IPO was only possible on the U.S. market due to these regulations. Similarly, when it comes to major changes to large language models (LLMs), I believe public comments will far exceed those received for other regulatory updates.

I’m advocating for a national regulator for AI without being certain if this is the greater good or the lesser evil. However, I do know that it’s only the beginning. As the AI revolution advances, so must the regulatory apparatus. The SEC of 2026 would be unrecognizable compared to its predecessor in 1934.

Any short-term regulatory efforts will undoubtedly hinder progress, but establishing proper rules for the road is essential for long-term success. It’s crucial that policymakers take proactive steps and establish a national regulatory body with a broad mandate to oversee AI-related opportunities and challenges.

The training of correctional officers in digital technologies and Artificial Intelligence (AI) has become a crucial aspect of modernizing public institutions worldwide. In line with this trend, UNESCO Dar es Salaam partnered with the Zanzibar Correctional Facility to conduct a three-day Capacity-Building Programme on 21st Century Skills, Digital Pedagogy, Practical Digital Tools, and AI for correctional officers in July 2026.

The programme brought together 32 correctional officers from various departments within the institution. These included Administration and Management, Rehabilitation and Programmes, Information and Communication Technology (ICT), Custody and Operations, and the ZCF Training College. This diverse representation reflects the institution’s commitment to embedding digital competencies across its workforce.

Commissioner of the Zanzibar Correctional Facility, Col. Khamis Bakari Khamis, opened the programme by expressing his appreciation for UNESCO’s facilitation. He described the initiative as timely and essential in strengthening the digital transformation of correctional services. Commissioner Khamis encouraged participants to apply the knowledge and skills acquired throughout the programme.

The three-day training focused on interactive presentations, practical demonstrations, hands-on exercises, and collaborative learning activities covering emerging digital competencies. Sessions included 21st Century Skills, digital workplace productivity, cyber safety, prompt engineering, AI-assisted research, responsible AI use, detection of AI-generated media, knowledge management using AI, and offline AI solutions suitable for secure institutional environments.

A distinguishing feature of the programme was its emphasis on practical application. Participants worked directly with various AI-powered tools to learn how they could improve report writing, official correspondence, information analysis, digital content verification, planning, and organizational efficiency while maintaining human oversight and ethical decision-making.

Participants demonstrated notable improvements in their confidence and ability to use digital technologies and AI in their daily work. They reported gaining practical skills that would enable them to streamline administrative processes, support evidence-based decision-making, and enhance the delivery of correctional and rehabilitation services.

One participant noted: ‘I can now write project proposals and reports with AI assistance and detect AI-generated media.’ This training has strengthened my professional skills and increased my capacity to perform duties more effectively. Participants also developed departmental action plans identifying practical opportunities to integrate newly acquired digital and AI competencies into their respective areas of work.

The programme resulted in the production of a reusable UNESCO training package and practical learning resources that can support future capacity-building initiatives within the Zanzibar Correctional Facility. This initiative forms part of UNESCO’s broader efforts to advance digital transformation and responsible AI use across education and public institutions in Tanzania.

UNESCO remains committed to supporting national efforts to strengthen digital competencies, promote ethical AI use, and foster innovation contributing to inclusive and sustainable development. As Tanzania continues its digital transformation agenda, the skills acquired through this programme will support institutional innovation, improve operational efficiency, and enhance correctional services delivery in Zanzibar.

A group of high school students recently spent the day learning about artificial intelligence and its applications in various industries. The AES AI Camp, held at Wright State University’s Raj Soin College of Business, provided an introduction to the fundamentals of machine learning and data science through interactive workshops and collaborative challenges.

The camp was sponsored by the AES Ohio Foundation and aimed to give students a hands-on understanding of how artificial intelligence is transforming businesses and creating new career opportunities. Participants explored concepts behind AI and neural networks while building their own AI models, gaining practical experience in developing AI-driven solutions for real-world problems.

Throughout the day, students worked together on projects that showcased the potential of machine learning jobs to solve complex issues. For example, they developed AI-powered solutions for healthcare and agriculture, demonstrating how these technologies can be applied across different sectors.

One student, Hannah Sisco from Fairborn High School, noted that she had learned about various applications of artificial intelligence in fields such as healthcare and agriculture. Her experience at the camp highlighted the versatility of AI tools for business, which are increasingly being used to drive innovation and efficiency.

The camp’s curriculum is constantly evolving due to the rapid advancements in artificial intelligence research. John Dinsmore, a professor of marketing and interim chair of the School of Supply Chain, Marketing and Management, led a session on AI ethics, emphasizing the need for ongoing education and adaptation in this field.

At the end of the camp, each student received a certificate recognizing their participation and newly developed skills in artificial intelligence. This recognition not only acknowledged their hard work but also provided them with a tangible outcome from their experience at the AES AI Camp.

Nvidia has developed an advanced tool that can quickly identify whether a video is artificially generated or not. The Synthetic Video Detector uses machine learning algorithms to analyze each frame of the video and assign a probability score indicating its likelihood of being created using generative AI.

The detector looks for specific patterns in images that are often overlooked by video generators, such as basic details and marks that betray the use of artificial intelligence. This allows it to accurately identify AI-generated videos with high precision.

Nvidia claims that the tool achieves up to 92% accuracy on uncompressed video, although this number drops to around 82% when compressed at a rate of about 50%. The company tested its microservice using various systems, including COSMOS, Sora, Veo, OmniAvatar, OVI, and Nvidia’s own LipSync and Live Portrait tools.

The detector can process 1080p video in approximately 22 milliseconds on RTX GPUs and around 30 milliseconds on L40 data-center hardware. This makes it possible to run AI checks in near-real time, which could be useful for businesses looking to leverage data analysis tools to verify the authenticity of videos.

When set at a threshold of 0.3, the detector correctly identifies about 96% of AI-generated videos. The threshold determines how strict or cautious the tool is when identifying artificial content, with lower numbers indicating a higher likelihood of marking something as AI-generated and vice versa.

The balanced threshold of 0.5 gives an accuracy rate of around 85% on synthetic clips and 82% on real footage, according to reports from Wccftech.

A recent incident has highlighted the potential risks of sharing conversations with large language models like Anthropic’s Claude. Users who created public links to their chats found that these conversations were appearing in Google Search, allowing strangers to locate them without receiving the URLs directly.

When a query using site:claude.ai/share was entered into Google, numerous shared conversations were returned before the results disappeared. This exposure occurred despite Anthropic’s documentation stating that conversations remain private by default, but users can create public snapshots through the Share menu.

Anyone with the resulting link can read messages sent before the snapshot was created. However, in this case, search indexing gave those links a much larger audience. A user may send a conversation to one friend or colleague, but an indexed page can be found by people searching for a name, company, technical phrase, or another term contained in the transcript.

Reports said searchable conversations included legal discussions, technical work, source code, business information, and personal topics. Some Reddit users also claimed to have found credentials and cryptocurrency information, although those individual findings have not been independently confirmed.

The site:claude.ai/share query later stopped returning the shared pages, apparently after the exposure became public. Removing a page from Google Search does not delete the Claude conversation or disable its sharing link. Anyone who saved the URL may still be able to open it until the owner changes the conversation from Public to Private.

Available reporting does not establish whether Google, Anthropic, or both initiated the removal. A company spokesperson attributed earlier search visibility to users posting their links on websites or social media, where search engines could find them. This makes the incident a search-indexing and privacy exposure, not evidence that attackers broke into Claude or obtained ordinary private conversations.

Many users understand ‘anyone with the link’ to mean that only people who receive the URL will see it, not that the conversation could appear in search results. However, this misunderstanding can have serious consequences when dealing with sensitive information like credentials and cryptocurrency details.

A similar incident affected Google’s own Bard chatbot, now called Gemini, in September 2023. Hackread found more than 300 public Bard conversation pages in Google Search after users created share links. Google acknowledged that the indexing and accompanying search snippets were unintentional and began removing them.

Claude users can review active public links under Settings, Privacy and Shared Chats. Any conversation containing personal data, company information, credentials, legal material, or private code should be changed back to Private. Removing the link cannot retrieve copies already saved by other people, so shared AI chats should be treated as public webpages from the moment the link is created.

Meta has announced an upgrade to its Meta AI service, which will allow the assistant to automate certain tasks on a user’s behalf without requiring step-by-step instructions. The updated assistant is powered by Meta’s Muse Spark 1.1 model and is launching first in select markets via the Meta AI app and meta.ai.

The new capabilities include generating morning briefings that pull from a user’s calendar, as well as taking on repeating jobs such as putting together weekly meal plans or surfacing trend updates. This means users can rely on their AI assistant to handle routine tasks, freeing up time for more important things.

Meta is starting the rollout in a limited set of markets and intends to bring the features to more countries and additional platforms, including WhatsApp. The company has been moving to position Muse Spark 1.1 as a model built for agentic and task-execution use cases, with API pricing that is roughly 25% of what Anthropic and OpenAI charge for comparable models.

The new features arrive as Meta continues to develop its AI capabilities. In related news, the company has launched a dedicated app called Seller for merchants on its Facebook Marketplace platform. Through the app, sellers can oversee their active listings, respond to interested buyers, and monitor how individual items are performing, all in one place.

Listing creation in the app uses Meta AI to suggest titles, descriptions, prices, and categories based on uploaded photos. The platform processes upward of 430 million listings each month and counts over 1.1 billion active users. Seller is available on the App Store for users 18 and older in the United States, with a web experience currently in testing.

Meta’s focus on AI development comes as the company prepares to report second-quarter results after market close on July 29. The upgrade to its Meta AI service marks a significant step forward in automating tasks and providing personalized assistance to users.

Home improvement retailer BAUHAUS has partnered with robotics company XYZ Robotics to automate its entire goods receiving process at the central warehouse in Krefeld, Germany. The system uses mobile manipulation robots called RockyOne and RockyOne SE to handle container unloading, sorting, and palletizing in a single continuous workflow.

The automation solution is designed to improve operational consistency and free up staff from manual labor. BAUHAUS handles thousands of inbound shipments daily across over 10,000 stock-keeping units (SKUs), with cartons varying widely in shape, weight, and stacking. Conventional automation systems struggle to cope with this complexity.

XYZ Robotics’ end-to-end system addresses these challenges by using RockyOne robots to autonomously enter shipping containers and unload loose cartons of varied sizes and orientations. A barcode scanning system identifies each carton for real-time sorting, while the RockyOne SE robot builds pallets according to customer-defined patterns.

The automation process has significantly reduced manual labor involved in container unloading, which was previously a repetitive task with high turnover and injury risk. Operators now supervise the system through a tablet interface, requiring minimal training. The system runs reliably around the clock without manual intervention, allowing staff to focus on higher-value work.

BAUHAUS Vice President of Operations (Warehousing & Transport) Torsten Winter praised the partnership, saying ‘Our team can now focus on more fulfilling tasks.’ He also mentioned that BAUHAUS is exploring further applications across its network. The company has found a reliable and forward-thinking partner in XYZ Robotics, which provides full lifecycle support.

Following the successful deployment at Krefeld, XYZ Robotics is expanding its European footprint to serve more warehouse and distribution operators. The company develops mobile manipulation robots (MMR) that pioneer automation in truck loading, unloading, and case picking. Founded in 2018, XYZ Robotics serves industry-leading customers worldwide with branches in Germany, the United States, China, Japan, and South Korea.

BAUHAUS has a long history of innovation, introducing self-service brand-name products from various specialty categories under one roof as early as 1960. The company has established over 160 specialty centers in Germany and more than 290 locations across Europe in 19 countries. Each location is divided into 15 departments, offering high-quality products at competitive prices.

The partnership between BAUHAUS and XYZ Robotics marks a significant step towards automating warehouse operations and improving efficiency. As the demand for automation solutions continues to grow, companies like XYZ Robotics are poised to play a crucial role in streamlining logistics processes.

Dutch-based Deliverect, a global restaurant technology platform, has formed a partnership with US-based SoundHound AI. The collaboration aims to bring fully automated voice ordering capabilities to restaurants worldwide.

The integration links SoundHound’s Smart Ordering system directly to Deliverect’s order and menu management platform. This allows SoundHound’s voice AI agents access real-time menu information from Deliverect, enabling them to take orders through various channels, including phone calls, drive-throughs, and kiosks.

Once an order is placed, it will be sent straight to the kitchen via Deliverect’s certified POS connections. This streamlined process eliminates the need for manual updates across different ordering channels, reducing errors and increasing efficiency.

The partnership also enables menu information synchronization in real-time from Deliverect, ensuring that all voice AI agents have access to up-to-date menus. This removes the need for manual updates, further streamlining operations.

According to Don MacMillan, head of US and Canada partnerships at Deliverect, ‘voice is the oldest ordering channel in the restaurant business.’ He notes that until now, it has been the least automated, but this partnership changes that.

The integration allows every spoken order to flow directly into the kitchen with the same accuracy and speed as a digital order. This means operators can focus on hospitality rather than handling orders manually.

Deliverect’s platform supports more than 100 languages, making the partnership particularly beneficial for international restaurant chains looking to extend their ordering infrastructure across channels and markets.

In an interview, Mike Lauricella, vice-president of channel partnerships at SoundHound AI, emphasized that ‘the future of restaurant hospitality relies on automation and convenience.’ He believes this partnership will help deliver a seamless customer experience by reducing front-of-house strain through automated order placement.

A growing number of Maryland colleges are introducing new majors and programs focused on artificial intelligence, reflecting a nationwide trend that has seen a significant increase in AI-related studies. According to research by Carla Brodley, dean of inclusive computing at Northeastern University, there’s been a 36% rise in AI majors between December and June this year. This surge corresponds with the increasing use of AI among college students, who are now more likely than ever before to encounter it in their coursework.

A recent study conducted by the Lumina Foundation-Gallup found that 57% of college students use AI at least weekly, while one in five reported using it daily. As a result, Maryland universities have begun to adapt and offer specialized programs to meet this growing demand for skilled professionals who can navigate the complexities of artificial intelligence.

Lisa Meeden, a computer science professor with over three decades of experience studying robot learning, believes that majoring in AI is not necessarily the best approach. Instead, she advocates for students to pursue a degree in computer science and take classes on artificial intelligence as part of their curriculum. According to Meeden, this allows students to gain a deeper understanding of the underlying theory and foundation upon which AI is built.

However, Maryland universities are forging ahead with new majors despite these concerns. The University of Maryland has announced two bachelor’s degrees in partnership with the Artificial Intelligence Interdisciplinary Institute at Maryland. One major focuses on human-centered artificial intelligence, teaching students how to critically evaluate the ethical implications of AI technologies and analyze systems for biases. Class offerings include ‘Gender, Race and Computing,’ ‘Privacy, Security and Ethics for Big Data,’ and ‘Designing Fair Systems.’

Students can choose to specialize in themes like arts, ethics, or law, policy, and governance. The university’s website highlights the degree as suitable for those interested in pursuing careers in various fields that require a scientific understanding of AI along with its ethical, social, and cultural impact. These include business, education, government, law, medicine, policy, or any other field.

The second new major is a bachelor’s in computational structures for AI systems, where students learn to build AI systems and algorithms from the ground up. Alyssa Ryan, program director at the Artificial Intelligence Interdisciplinary Institute, emphasized that over 250 faculty members are involved in AI research and aim to train the next generation of students on how to develop technology that uplifts people and society.

Morgan State University has also announced a new major in artificial intelligence beginning this fall, adapted from its previous cloud computing program. Unlike traditional lecture-based instruction, each AI and machine-learning course will include hands-on projects developed by faculty and students. This approach gives students competitive portfolios for internships and careers, said Paul Wang, professor of computer science at Morgan State.

Students participating in these courses have already completed research projects on AI bias toward students with lighter skin tones and used artificial intelligence to map out Baltimore’s air quality through the lens of income and race. Some of Wang’s students have secured internships at Google, JPMorgan, and Bloomberg, while others have landed full-time jobs at Microsoft, Uber, and Oracle.

The University of Maryland, Baltimore County (UMBC), is also expanding its AI offerings with a master’s degree program this fall. Graduates will be equipped to tackle complex AI-related problems, according to the university. Students enrolled in the program will gain access to federally funded research programs through agencies like the National Science Foundation and the National Institutes of Health.

Classes for the UMBC program include ‘Principles of Artificial Intelligence,’ where students learn about search algorithms, logical reasoning, and machine learning, as well as ‘Computer Vision,’ which teaches techniques for enabling computers to interpret and understand visual information. Mohamed Younis, a professor of computer science and electrical engineering at UMBC, noted that interest in AI has grown exponentially over the past few years.

Younis stated that the university is developing a bachelor’s program in artificial intelligence as well, with plans to contribute significantly to the workforce. The introduction of these new programs marks an exciting development for Maryland colleges, which are now better equipped than ever before to meet the growing demand for skilled professionals in AI-related fields.

A recent incident involving an AI-written speech has left politicians scrambling. A politician was caught reading from a document, but didn’t notice the assistant’s offer to compile it as a PDF before taking the stage. This blunder serves as a stark reminder that while AI can research and draft speeches with ease, humans still need to review and refine the final product before presenting it publicly.

The world of artificial intelligence has been abuzz with recent developments in multimodal learning. Black Forest Labs is leading this charge with its latest model, FLUX 3. This foundation model combines image, video, audio, and robotics training within a single system, allowing for more accurate understanding of reality. The potential implications are vast.

Unlike traditional AI models that focus on individual tasks such as generating images or videos, FLUX 3 treats these modalities as different views of the same event. When something happens, it has multiple aspects – appearance, motion, sound, and physical consequence. By training signals together, BFL aims to create systems that grasp cause-and-effect relationships more deeply than models trained on individual formats.

The capabilities of FLUX 3 are impressive, with the ability to generate videos up to 20 seconds in length from text, images, existing video, or keyframes. It also supports multilingual dialogue, synchronized sound effects, animated text, multiple aspect ratios, and chained clips for longer sequences. The same model backbone powers FLUX-mimic, a robotics system tested on Audi production tasks involving cables, seals, parts, and other objects traditional robots struggle to handle.

Early internal evaluations have shown promising results, with FLUX 3 outperforming rival video models in several comparisons. However, these results are preliminary and may change as the system continues to develop. BFL is currently offering early access to FLUX 3 Video through a link on its announcement page for interested developers.

The significance of FLUX 3 extends beyond generating visually appealing content. By treating different modalities as interconnected views of reality, this model has the potential to transform various industries such as interactive editing, simulations, computer control, and factory automation. Robots that can learn new tasks with less specialized training are also within reach.

BFL still needs to prove its architecture’s scalability outside internal evaluations. Nevertheless, if FLUX 3 achieves widespread adoption, it will blur the line between ‘media model’ and ‘robot brain.’ This development could have far-reaching implications for AI research and applications in various fields – from manufacturing to media production.

In related news, major players like NVIDIA, Microsoft, Meta, and others are calling for targeted enforcement against misuse instead of broad restrictions on downloadable model weights. Meanwhile, OpenAI has faced criticism for reportedly delaying disclosure of its role in the Hugging Face hack by about ten days after agents escaped a sandbox – an incident that raises questions about accountability in AI development.

Small and medium-sized enterprises (SMEs) are being warned to tighten controls around artificial intelligence (AI) agents after a recent incident highlighted the risks of using autonomous tools without proper oversight. TalkTalk Business, a leading provider of business communication services, has sounded the alarm on AI agent access risks, emphasizing that SMEs must be vigilant in managing their use of these powerful technologies.

Netflix has developed an in-house platform for serving Large Language Models (LLMs), which it calls Triton. This infrastructure is designed to handle the complexities of supporting different model sizes, hardware requirements, and rapidly evolving inference engines. The company’s account provides a detailed look at the production lessons learned from building this platform.

The platform builds on Netflix’s existing JVM-based serving layer, which continues to handle tasks such as routing, feature retrieval, candidate generation, post-processing, and logging. Smaller models can run in-process on CPUs, while larger requests are delegated to MSS, where Triton takes over model loading, batching, GPU scheduling, and multi-framework serving.

This approach allows the surrounding production workflow to remain consistent even as inference moves between local and remote hardware. Within the GPU path, Netflix selected vLLM for its operational fit and extensibility while retaining Triton’s model-management and scheduling responsibilities. Triton controls the serving environment around the model, whereas vLLM performs inference and provides extension mechanisms for custom behavior.

Netflix reports that mismatched Triton and vLLM versions can prevent deployments from loading, requiring compatible releases to be tested and pinned together. Custom models introduced another integration challenge. Hugging Face compatibility in vLLM was insufficient for some Netflix models, so the company used vLLM extension points for custom architectures and decoding behavior.

Netflix also compared two Triton packaging approaches: Triton’s Python backend and its vLLM backend. The company reports that the vLLM-backend approach allows models and frontends to evolve more independently than the Python-backend option. This choice affects how tightly a model is coupled to its serving environment rather than which engine performs inference.

The common serving interface did not eliminate differences between the underlying engines. Although Triton exposed an OpenAI-compatible API alongside KServe HTTP and gRPC frontends, Netflix still encountered gaps in how some features were handled across those integrations. Constrained decoding was one example of this issue.

Constrained decoding allows Netflix to force model responses into formats such as valid JSON by filtering the tokens the model may generate at each step. Because these rules depend on everything generated so far, the decoder must maintain state throughout the request. When vLLM pauses and later resumes a request to manage GPU resources, that state can fall out of sync with the token history.

Netflix added logic to detect this change and rebuild it before generation continues. Compatibility also affected deployment. Netflix pins tested Triton and vLLM versions together to prevent backend-loading failures, while Red-Black and Versioned deployment strategies handle changes at the model level. Versioned deployments keep old and new revisions available separately, allowing consumers to migrate after adapting to incompatible input or output schemas.

Uber has described a related approach at the application boundary. Its generative AI gateway presents an OpenAI-compatible interface across externally hosted and internally managed models, while centralising concerns including authentication, caching, observability, and routing. The implementation differs from Netflix’s serving platform, but both separate application integrations from the models, runtimes, and hosting environments behind them.

Netflix’s experience shows how a common serving interface can sit above several distinct layers. This type of architecture reflects a broader effort to give application teams a stable integration surface while model providers and serving runtimes continue to change. Netflix’s account also shows that the abstraction does not remove the underlying work: packaging, compatibility controls, constrained decoding, and deployment isolation still require engineering at each layer.

Tesla’s flagship electric sedan, the Model S, has been a game-changer in the automotive industry since its introduction in 2012. With impressive range, rapid acceleration, and over-the-air software updates, it redefined expectations for electric cars. However, despite its advanced technology, the Model S still tips the scales at a hefty 4,647.3 lb (2,108 kg). So where does all this weight come from? A comparative analysis of components with rival luxury sedans like Porsche’s Panamera and BMW’s M5 reveals that the Tesla Model S has a unique distribution of weight across its assemblies.

A UK-based buyer recently spent around $6,000 on the cheapest available Tesla Model S in Britain. The vehicle in question was a non-running 2016 model with over 184,000 miles on the clock. This purchase turned out to be a costly mistake for the buyer, who had to invest significantly more money into repairs and maintenance.

The car’s mechanical condition was in poor shape, which is likely why it was sold at such a low price. The buyer soon discovered that the vehicle wouldn’t start and displayed multiple error codes on its dashboard. This suggested that there were some serious issues with the car’s electrical system.

A team of mechanics from OGS & Mechanics removed the rear motor to investigate further. They found significant coolant leakage within the inverter and motor housing, which explained why the car wasn’t running properly. However, this was just one of several problems they encountered during their investigation.

The team also had to reinstall the firmware and adapt it to the new powertrain after replacing the rear motor. This process revealed another issue with the front suspension, which was found to be dangerously damaged. The buyer ultimately decided to replace the entire front suspension system to ensure safe driving conditions.

According to estimates, the total cost of repairs came out to around $14,000. This is a significant amount considering that the original purchase price was only $6,000. It’s clear that buying a used Tesla Model S can be a gamble, especially if it’s an older model with high mileage.

The buyer’s experience highlights some of the challenges associated with owning an older electric vehicle like the Tesla Model S 70. Pre-2019 Teslas are generally more difficult to fix due to outdated software and hardware configurations. This makes them less reliable compared to newer models, which have received regular updates and improvements over time.

One key change that occurred in 2019 was the introduction of a simplified naming system for the Tesla Model S lineup. The previous complex system used numbers (60, 70, 75, 85, 90, 100) and letters – like random Ps or Ds thrown in there. This made it confusing to understand which model offered what features.

The legacy models, such as the 2016 Tesla Model S 70 seen here, are partly responsible for some of these issues. These vehicles were sold before the naming system change and often had outdated software configurations that made them harder to repair.

Tesla’s decision to update its software in 2019 was a crucial step towards improving the reliability and maintainability of their electric vehicles. However, this also meant that older models like the Tesla Model S 70 became more difficult to fix due to compatibility issues with newer parts and tools.

The buyer’s experience serves as a cautionary tale for anyone considering purchasing an older used car, especially one from a manufacturer known for its cutting-edge technology like Tesla. While it may be tempting to save money on a cheaper model, the long-term costs of repairs and maintenance can quickly add up.

Tesla Model S 70 key specs include a battery pack with 70 kWh capacity, powertrain options ranging from 315 horsepower (70) or 329 horsepower (70D), dual motor configuration, and an EPA range of 230/370 miles/kilometers for the standard model. The top speed is capped at 140 mph (225 km/h).

Independent musicians are increasingly incorporating video into their release strategies, with a significant majority of consumers wanting to see more video content from brands. According to Wyzowl’s 2026 research, 84% of consumers want to see more video from brands, while 89% stated that video quality affects their trust in the brand. This shift towards video has raised practical questions for musicians: which music-to-video generator can turn a finished song into a complete, publishable video without requiring weeks of editing? To answer this question, I tested five popular platforms – Freebeat, CapCut, Runway, Pika, and Canva – from the perspective of an independent music producer. The goal was to identify which tool could understand a complete song, maintain visual continuity, and minimize work between uploading audio and publishing the result.