AI Safety news and analysis on Web Pulse: 12929 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A viral hoax misattributed a fictional AI model named Le Chaton Fat to Mistral AI, after Mistral itself posted and deleted a satirical announcement for a 24-trillion-parameter model. Community memes inflated the joke wit…
France's DGSI intelligence agency replaced US data-analytics firm Palantir with French startup ChapsVision on a major contract, Prime Minister Sebastien Lecornu announced June 16, 2026, citing the need to avoid strategic…
Anthropic disabled access to its most advanced AI models, Fable 5 and Mythos 5, for all customers after a Trump administration directive ordered the company to block foreign nationals from using them. The company complie…
Anthropic released Fable 5, a highly anticipated AI model, on June 9th, 2026, after reports of dangerous capabilities. A user testing the model found it good but not mind-blowing, noting improvements in reasoning and cod…
The U.S. government shut down global access to Anthropic's most powerful AI models, triggering a global scramble for sovereign AI as countries realize their dependence on U.S. technology is a strategic vulnerability. Eur…
Authorizer 2.3.0 introduces Zanzibar-style fine-grained authorization (FGA) for vector search, enabling permission-aware RAG by pre-filtering document retrieval based on user access rights. The update includes an MCP ser…
A Hacker News user proposes using hardware passkeys like YubiKey for short-lived access token creation in CLI tools for AI services and GitHub, arguing it would enhance security by limiting token lifespan and reducing at…
The US Army commissioned three more tech executives, including Cloudflare CTO Dane Knecht, into Detachment 201 as lieutenant colonels, bringing the total to seven Silicon Valley leaders advising the Pentagon on AI and mo…
Microsoft researchers propose using information-flow control (IFC) to secure autonomous AI agents against prompt injection and data exfiltration. IFC labels data with integrity and confidentiality tags, propagates them t…
Flowork introduces a centralized settings panel for managing API keys, global defaults, and agent configuration. The owner-level control panel stores credentials in a global database, injects them into running processes …
AI code scanners become trustworthy only when fed explicit threat models, according to a cross-domain analysis of autonomous systems. Production agents require distributed-systems rigor including idempotent tool calls, c…
An 86-year-old man in Seattle uses an AI-enabled microphone called Sensi.ai to monitor his movements and detect falls, allowing him to age in place while his son monitors from Austria. The device records private conversa…
TekStream launched TekStream Proactive Cyber Defense, an expert-operated security service powered by its Cosmos platform, to counter AI-driven threats including autonomous attacks. The service addresses a threat landscap…
Anthropic, the AI safety company that spent six months warning about the technology's risks, saw the White House shut down its flagship models after the Pentagon designated it a supply chain risk and the company filed fo…
Vallejo police arrested an Antioch man attempting to carry a concealed firearm into the Vallejo High School graduation at Corbus Field, just eight days after a fatal shooting at a Fairfield graduation. Officers recovered…
Over 65 local leaders gathered in Saratoga on June 11 to discuss leadership during emergencies, focusing on communication, trust, and community support. The event, organized by Santa Clara County Cities Association Presi…
Emergency veterinarians in Marin County report a dramatic spike in dogs suffering from suspected mushroom poisonings, with over 10 cases treated in three weeks. The death cap mushroom (Amanita phalloides), which causes s…
Code strings in iOS 27 beta suggest Apple plans to add a break reminder for long Siri AI conversations, warning users that Siri is not a real person. The feature aims to address risks of parasocial attachment to AI, as A…
KU Leuven's Centre for IT & IP Law will host a conference on 10 December 2026 in Leuven to examine the risks and opportunities of the EU's Digital Omnibus proposal, which aims to simplify digital regulations across AI, d…
KU Leuven's Centre for IT & IP Law (CiTiP) will host a conference on 10 December 2026 in Leuven to examine the EU's Digital Omnibus proposal and its impact on democracy, digital sovereignty, AI fairness, and other areas.…