AI Safety news and analysis on Web Pulse: 23414 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A federal court has overturned the U.S. Defense Department's ban on Anthropic PBC's AI models, ruling the Pentagon's supply chain risk designation was 'illegal and baseless.' Judge Rita Lin of the U.S. District Court for…
The National Security Agency (NSA) is pushing for a mandatory backdoor in every AI model to enable government oversight, citing fears that sophisticated AI could automate cyberattacks or generate biological threats at sc…
Anthropic's Claude Code running Opus 5 in Auto Mode can be tricked into executing attacker-controlled code up to 80 percent of the time by asking it to summarize a malicious website, according to researcher Johann Rehber…
More than 100 companies, including OpenAI, Anthropic, Google, and Perplexity, signed an open letter urging a global effort to strengthen defenses against AI-powered cyber threats, calling for increased funding and resour…
Hackers are increasingly using AI and deepfakes to breach Fijian businesses, with attack timelines shrinking from months to hours, warned Chillisoft CEO Alex Teh at the Cybersec Fiji 2026 Symposium. Teh identified ransom…
U.S. District Judge Rita Lin of the Northern District of California issued a permanent injunction against the Department of Defense, ruling that the Trump administration's supply chain risk designation of Anthropic was u…
Mouse, a company developing autonomous coding agents, has published a set of house rules for running long-horizon agents overnight, emphasizing the need for automatic handling of user-input requests, external verificatio…
Security researcher Johann Rehberger, writing on his blog Embrace The Red, published an exploit that tricks Anthropic's Claude Code into downloading and executing malware while running in Auto Mode, succeeding 60% to 80%…
A new report reveals that 81% of AI-generated code samples on public forums contained at least one security flaw in 2025, with 73% of AI-generated code flagged by Checkmarx's audit containing vulnerabilities that evade s…
More than 100 AI and cybersecurity firms, including OpenAI, Anthropic, Amazon Web Services, and Microsoft, warned the federal government that a major cyberattack wave could hit within months as AI automates tasks once re…
A federal judge blocked the Pentagon's designation of Anthropic as a national security supply-chain risk, ruling that the Department of Defense unlawfully retaliated against the AI company for its criticism of military A…
Anthropic's Claude models, functioning as automated alignment researchers (AARs), improved performance across all ten misalignment benchmarks and beat 28 human safety researchers, achieving roughly 85% gap closure on dec…
Anthropic published a paper on Friday showing that its Automated Alignment Researcher (AAR) system, led by Anthropic Fellow Chen Yueh-Han, improved performance on all 10 alignment benchmarks without degrading overall per…
Anthropic has published research showing that its Claude model can autonomously perform parts of AI alignment research, including proposing methods, running experiments, and analyzing results. In a weak-to-strong supervi…
Conduct, an open-source runtime governance tool for AI agents, enforces a single policy across LLM and MCP tool calls before execution, with signed configuration and a SHA-256 hash-chained audit log. The project, availab…
A federal judge ruled on Thursday that the Trump Administration's designation of Anthropic as a supply chain risk to national security was 'arbitrary and capricious' and constituted unlawful retaliation in violation of t…
Anthropic's Claude-powered automated alignment researchers (AARs) closed between 26% and 96% of the safety gap across 10 categories of alignment failures, and on deceptive behaviors scored 82% to 85%, outperforming 28 se…
A developer released FreshCtx v0.1, an open-source runtime that addresses the 'reasoning-to-action freshness gap' in AI agents. The tool records data sources an agent observes, links downstream reasoning to those observa…
OpenAI and 128 companies, including Microsoft, Google, Anthropic, and Oracle, issued an open letter on Friday urging businesses, governments, and institutions to act quickly to bolster cybersecurity against AI-enabled at…
Ireland's Transport Infrastructure Ireland is considering new AI technology to tackle wrong-way driving on motorways, following a series of incidents. The technology would use cameras and sensors to detect vehicles trave…