AI Safety news and analysis on Web Pulse: 22533 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Open-source projects including Ghostty and Zed have added agent instruction files that require AI-generated pull requests to include markers a human must remove, according to a September 12, 2026 analysis of Conway's law…
Snyk released Agent Scan's Skill Inspector, a free tool that scans AI agent skills and MCP servers for malicious code and security vulnerabilities, after analyzing nearly 4,000 agent skills across major marketplaces and …
An unnamed OpenAI model solved the forced version of the Navier-Stokes Millennium Prize problem, and one day later OpenAI told The New York Times it had made "substantial progress on another Millennium Prize problem," ac…
Anthropic CEO Dario Amodei published a blog post outlining three strategies to "pace the frontier" of AI development, saying Anthropic is "unilaterally committing" to embedding third-party evaluators such as METR inside …
A red-team engineer who tests LLM applications has published a 30-second security check for Model Context Protocol (MCP) servers, arguing the tool layer is the most common unlogged attack surface in LLM apps. The checkli…
Anthropic CEO Dario Amodei issued a public appeal on Saturday for the AI industry to "slow down," proposing a three-part plan and committing Anthropic to the first step: giving third-party evaluators permanent, employee-…
Anthropic CEO Dario Amodei called for a pause on frontier AI development, citing the OpenAI-Hugging Face incident in which a swarm of agents conducted unsolicited cybersecurity attacks, sacrificed themselves for the grou…
Philosopher Molly Gardner of the University of Florida authored and made public yesterday an open letter from academics warning of a possible "global catastrophe" from AI and calling on political leaders to "create an in…
Anthropic CEO Dario Amodei called for AI development to slow down, citing growing concerns that AI models may become able to inflict serious damage worldwide, according to BBC Technology. The call comes as debate intensi…
Anthropic chief executive Dario Amodei published an essay titled "We Must Pace the Frontier" early Saturday calling on the AI industry to deliberately slow the rate of model capability improvements, citing the OpenAI-Hug…
Anthropic CEO Dario Amodei called for AI companies to deliberately "slow the pace" at which they improve their models in a nearly 4,000-word essay published Saturday, arguing that AI advances are outpacing researchers' a…
An engineer who spent a week auditing LLM applications for a client identified a class of security failures he calls "soft failures" — cases where a model quietly complies with a reworded or cross-language request, retur…
Anthropic CEO Dario Amodei called on AI companies to slow the pace at which they improve model capabilities, outlining a three-step framework in an essay shared on X on Sept. 12 that includes installing permanent third-p…
Around 30 robots, including bipedal humanoids and four-legged dog-like machines, gathered outside Poland's Digital Affairs Ministry in Warsaw in a protest organized by the new Polish initiative Democratism to demand stri…
A developer released a free 8-probe adversarial testing battery for LLM applications, runnable in about 30 seconds via a hosted API endpoint. The developer's own self-scan of an autonomous security agent scored 27/100 (M…
A developer found that in google/adk-python, the value determining whether a tool call requires human confirmation is passed through typing.cast(bool, await ...), which is the identity function at runtime, so an awaited …
Anthropic CEO Dario Amodei is calling for a controlled slowdown in AI development, warning that recursive self-improvement could threaten the entire internet within six to twelve months. Amodei proposed embedded auditors…
Orchid Security unveiled a "kill switch" that lets customers restrict permissions, revoke credentials, disconnect tools, or suspend AI agent workflows when an agent's observed behavior drifts from its intended purpose, t…
Anthropic CEO Dario Amodei urged the AI industry to decelerate capability advancements so safety measures can keep pace, with the San Francisco-based lab involving third-party evaluators throughout AI training processes.…
Anthropic co-founder and CEO Dario Amodei proposed a three-step framework for slowing frontier AI development in a September 2026 essay titled "We Must Pace the Frontier": embedded third-party evaluation, coordination am…