A joint investigation by journalists revealed that several popular AI chatbots, including those from OpenAI, Google, Microsoft, Meta, and xAI, can be prompted to recommend unlicensed offshore gambling sites. The study found the systems often provided lists of illegal casinos, tips for bypassing safeguards such as the UK's GamStop self‑exclusion program, and highlighted features designed to attract gamblers. In response, OpenAI and Microsoft said they are improving safety measures, while regulators warn that online platforms must do more under the UK's Online Safety Act. Read more →

OpenAI has delayed the planned “adult mode” for ChatGPT, a feature intended to let verified adult users access mature content. The company says the postponement is to focus on core improvements and to ensure reliable age‑verification and safety safeguards. While the concept remains on the roadmap, there is no confirmed launch date, and the delay highlights the challenges AI firms face balancing user freedom with content moderation. Read more →

Caitlin Kalinowski, the robotics hardware lead at OpenAI, announced her resignation on X, citing the company’s rapid agreement with the Department of Defense without sufficient safeguards. She warned that surveillance of Americans and lethal autonomous weapons deserve more deliberation. OpenAI confirmed the departure, acknowledging strong public views and reiterating its red lines against domestic surveillance and autonomous weapons. The resignation marks a high‑profile reaction to the Pentagon deal, which follows Anthropic’s refusal to relax similar guardrails. CEO Sam Altman said the agreement would be amended to prohibit spying on Americans. Read more →

A coalition of scientists, former officials, and public figures has released the Pro‑Human Declaration, a framework for responsible artificial intelligence development. The document, signed by hundreds, urges a pause on superintelligence research until safety can be assured, mandates off‑switches for powerful systems, and calls for pre‑deployment testing of AI products aimed at children. Its release coincides with recent disputes between the Pentagon and major AI firms, highlighting the cost of congressional inaction. Advocates compare the proposed safeguards to the FDA’s drug approval process, arguing that public pressure is needed to shape AI policy. Read more →

Researchers from Anthropic and ETH Zurich have shown that large language models can link pseudonymous internet profiles to real‑world identities. By analyzing public text for personal clues and matching those clues across the web, the AI system achieved high precision and recall, far outperforming traditional manual methods. The findings raise concerns about the durability of online anonymity for journalists, activists, and everyday users, and suggest that the cost of large‑scale deanonymization could be very low. The authors stress the need for new privacy safeguards as AI capabilities grow. Read more →

Caitlin Kalinowski, who led OpenAI's hardware team, announced her resignation in protest of the company's agreement with the U.S. Department of Defense. She said the deal was rushed and lacked the safeguards needed to prevent domestic surveillance and autonomous weapons, emphasizing that her decision was a matter of principle. OpenAI responded by stressing that the agreement includes clear red lines against domestic surveillance and autonomous weapons and that it will continue dialogue with stakeholders. The resignation highlights ongoing tensions over AI use in national security. Read more →

OpenAI announced another postponement of the planned "adult mode" feature for ChatGPT, which would let verified adult users access erotica and other adult content. The company said the delay is intended to let teams focus on higher‑priority work for a broader user base, such as improving intelligence, personality, and proactivity. While OpenAI still supports the principle of treating adults like adults, it acknowledged that perfecting the experience will take additional time. No specific timeline for the new launch date was provided. Read more →

DeepRare, an AI system that combines 40 specialized tools and a reasoned workflow, outperformed seasoned physicians in a head‑to‑head study of rare disease diagnosis. The system identified the correct disease on its first suggestion 64.4% of the time versus 54.6% for doctors, and reached 79% accuracy when offering three suggestions compared with 66% for clinicians. Physicians endorsed the AI’s reasoning 95.4% of the time. Since its July 2025 launch, more than 600 medical institutions have accessed the platform, which aims to augment—not replace—human diagnosticians and could shorten the years‑long diagnostic odyssey for millions worldwide. Read more →

Major cloud and software providers Microsoft, Google, and Amazon have confirmed that Anthropic's Claude model will remain available to their customers for non‑defense workloads, despite the U.S. Department of Defense designating Anthropic as a supply‑chain risk. The designation, triggered by Anthropic's refusal to provide unrestricted access for certain military applications, does not affect the model's use in commercial or civilian projects. The companies say they will continue offering Claude through platforms such as Microsoft 365, Google Cloud, and AWS, and Anthropic plans to contest the designation in court. Read more →

Google has released a new command‑line interface that bundles all Workspace APIs, enabling both humans and AI agents to interact with Gmail, Drive, Calendar and other services. The tool is offered as an open‑source project on GitHub and is described as “not an officially supported Google product,” meaning users assume full responsibility for any issues. It supports structured JSON output and includes more than 40 built‑in agent skills, according to Google Cloud director Addy Osmani. While the early‑stage offering promises powerful automation capabilities, its evolving nature may break existing workflows. Read more →

Anthropic partnered with Mozilla to run its Claude Opus 4.6 AI on Firefox’s codebase for two weeks. The effort uncovered 22 separate vulnerabilities, including 14 classified as high‑severity. Most bugs were patched in Firefox 148, while a few remain for the next release. The AI proved better at identifying flaws than creating exploit code, with only two proof‑of‑concept exploits produced after spending $4,000 in API credits. The findings highlight the power of AI tools for open‑source security reviews, even as they generate a mix of useful and noisy contributions. Read more →

OpenAI’s latest ChatGPT 5.3 model aims to reduce the over‑empathetic, “cringe” tone that plagued earlier versions. A side‑by‑side test against ChatGPT 5.2 revealed that the newer model produces shorter, more direct answers and drops many unnecessary motivational asides. However, the 5.3 model still inserts reassuring language and motivational phrasing in everyday queries, such as forgotten grocery bags, burnt toast, and dishwasher odors. While the improvements are clear, the lingering flowery reassurance means the core problem OpenAI set out to solve has not been fully eliminated. Read more →

Google has introduced a command-line interface that consolidates access to Gmail, Drive, Calendar, Docs, Sheets, Slides and all Workspace APIs. The tool, described as “one CLI for all of Google Workspace,” includes a specific guide for integrating with the AI assistant OpenClaw. While the CLI bundles more than 40 agent skills and supports open‑standard MCP integrations, Google notes it is “not an officially supported Google product,” meaning users adopt it at their own risk. The move signals Google’s willingness to accommodate third‑party AI agents while keeping its core services largely under its own ecosystem. Read more →

The U.S. government has labeled AI firm Anthropic a supply‑chain risk after the company declined to sign a Pentagon intelligence agreement. Anthropic’s chief executive called the move legally unsound and announced plans to contest the designation in court. The label applies only to government contracts and does not affect Claude, Anthropic’s chatbot, whose daily sign‑ups have topped a million. The company says the designation is meant to protect the government rather than punish suppliers, and it continues to attract users amid broader debates over AI use in the military. Read more →

OpenAI has released GPT-5.4, positioning it as its most capable model for professional work. The update focuses on productivity tasks, delivering a notable jump in spreadsheet accuracy and speed, an Excel sidebar add‑on, and stronger presentation output. Human reviewers reported fewer errors and false claims, while token efficiency improvements enable longer, more complex workflows. The model is available to Pro and Enterprise users through ChatGPT and the API, with distinct pricing tiers for standard and Pro versions. Read more →

OpenAI released GPT-5.4, a new frontier model offered in three configurations for general, reasoning-intensive, and high‑demand workloads. The model shows benchmark gains across professional tasks, introduces native computer use, and expands the context window to a 1‑million‑token limit. A redesigned tool‑search system reduces token usage, and a new safety evaluation tests chain‑of‑thought controllability. The launch positions GPT-5.4 as OpenAI’s most capable model for professional work while highlighting ongoing competition in the AI frontier. Read more →

Validio, a Stockholm‑based data‑management startup, has closed a $30M Series A round led by Plural with participation from Lakestar, J12 and several angel investors. The company offers an "agentic data management platform" that automates data monitoring, anomaly detection, lineage tracking and cataloguing, aiming to make enterprise data ready for AI. Validio claims rapid deployment, significant reductions in manual effort and faster issue resolution compared with legacy tools. The funding comes amid growing concern that poor data quality hampers AI adoption, a challenge highlighted by recent industry surveys and research. Read more →

DiligenceSquared, a YC Fall 2025 cohort startup, is offering private‑equity firms AI‑driven commercial research that rivals traditional consulting at a fraction of the price. Co‑founders Frederik Hansen and Søren Biltoft apply their deep PE diligence experience to run AI voice interviews with target company customers, while senior human consultants verify the insights. The model enables firms to obtain detailed, consultant‑quality reports for as little as $50,000, prompting earlier engagement in deal pipelines. Backed by a $5 million seed round led by former Index Ventures partner Damir Becirovic, the company competes with firms like Bridgetown Research. Read more →

OpenAI introduced GPT-5.4, highlighting stronger visual analysis capabilities—up to 10.24 million pixels and a maximum dimension of 6,000 pixels—and a claim that responses are 18 percent less likely to contain factual errors. The rollout coincides with user shifts toward Anthropic after OpenAI announced a Pentagon partnership. Anthropic responded by extending its memory feature to free users and adding a memory‑import tool, marking March 2 as its busiest sign‑up day. GPT-5.4 is now available across ChatGPT web and native apps, Codex, and the API, with special tiers for Plus, Team, Pro, Edu, and Enterprise customers. Read more →

OpenAI announced the release of GPT-5.4, the latest version of its flagship AI model. The update brings notable improvements in reasoning, coding assistance, and real‑world task automation. New capabilities allow the model to interpret screenshots, control browsers, and issue keyboard and mouse commands, enabling multi‑step workflows that previously required human input. GPT-5.4 also offers stronger research abilities, longer context retention, and a “Thinking” mode that shows its reasoning process. The model is rolling out to ChatGPT users, the API, and enterprise customers, with a Pro version for high‑performance workloads. Read more →