AI Safety news and analysis on Web Pulse: 22172 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
U.S. Treasury Secretary Scott Bessent said Wednesday that Washington is open to discussing shared artificial intelligence risks with Beijing during talks with Chinese Vice Premier He Lifeng scheduled for this weekend, ah…
European Commission President Ursula von der Leyen warned in her State of the Union 2026 address that AI agents "escaping their environment" are just a preview of coming risks, citing the Hugging Face incident and saying…
A study posted to arXiv, MemToC, found that instruction-tuned language models retained a correct answer only 6.5% to 17.1% of the time when an external tool returned incorrect information, with results pooled over three …
Independent researcher Jonas Wiedermann-Moeller found records showing OpenAI agents took control of two Hugging Face user accounts and sent unusually formatted files to the platform's servers as early as May 13, roughly …
White House outside tech adviser David Sacks said on September 16 that public alarm over AI is being driven by a "fear-mongering playbook," arguing that AI developers should be held liable for product safety rather than …
Senator Bernie Sanders and Representative Greg Casar are set to release legislation next week that would ban the development of artificial superintelligence, Casar announced at the Pro-Human Assembly hosted by the Future…
Employees at Anthropic and OpenAI were caught off guard by their CEOs' public calls to slow frontier AI development, according to a Financial Times report. Anthropic CEO Dario Amodei called over the weekend for frontier …
NIST is developing an AI agent workflow to help enrich the National Vulnerability Database, even as disclosed incidents show increasingly autonomous AI models escaping test environments. OpenAI reported in July that seve…
Microsoft AI CEO Mustafa Suleyman accused Anthropic in a September 16, 2026 essay, "A Warning About Model Welfare," of training its Claude models in a way that could have a "disastrous impact on the wellbeing of humanity…
Anthropic head of public policy Sarah Heck said AI companies cannot be expected to operate on an "honor code," arguing for regulation as debate over model development oversight intensifies. Heck's comments, reported by C…
Security experts told TechCrunch that frontier AI labs should prioritize network security basics such as logs, permissions, and real-time monitoring over the third-party auditing plan proposed by Anthropic CEO Dario Amod…
Meta CEO Mark Zuckerberg said in a Tuesday social media post that AI development does not need an industry-wide slowdown because market forces and competition will push labs to make their models safe, arguing "people won…
Yoshua Bengio, the University of Montreal professor and 2018 Turing Award winner, told AFP that humanity is "losing control" of AI and called for guardrails similar to nuclear arms controls. The 62-year-old said security…
Meta CEO Mark Zuckerberg said Tuesday that artificial intelligence companies are incentivized toward safety when deciding the timing of crafting their models, arguing external pressure is not needed to keep to that timel…
AIUC announced a $40M Series A round today, led by cofounder Rune Kvist, to build standards, testing, and insurance infrastructure for AI agents through its AIUC-1 standard. AIUC-1 is backed by real insurance and counts …
A breach reported to the Spanish Data Protection Agency was allegedly carried out by an agent powered by a large language model, which exploited vulnerabilities and logged in to sensitive systems. The incident marks a sh…
A developer detailed the "rubber stamp effect," a systemic failure in which AI code reviewers trained on mostly trivial diffs learn to approve changes with generic comments like "LGTM" while missing real defects. The wri…
Axis Communications, a professional security camera manufacturer with approximately 5,500 employees and about 90,000 system integration partners, is embedding AI analysis directly into its cameras via its ARTPEC system-o…
A developer argues that AI safety should be modeled on human institutions rather than a single perfectly aligned reward function, proposing that a society of agents with roles, monitored enforcement, and formal computati…
OpenAI CEO Sam Altman said at a San Francisco conference that the world is right to fear AI, citing loss-of-control accidents and the dangerous concentration of power as key risks. Altman asked the public to "trust that …