AI Safety news and analysis on Web Pulse: 22400 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A new arXiv paper (2609.13672v1) introduces recoverability as a system primitive for long-horizon AI agents, making the reuse of saved state an explicit decision rather than an automatic continuation. The paper's referen…
A new arXiv survey (2609.13731v1) synthesizes 206 foundational studies and regulatory standards into a systems-security reference framework for trustworthy agentic AI, formalizing the general agent architecture as a stat…
A new arXiv paper (2609.13552v1) proposes the AI Trust and Assurance Layer (ATAL), a model-agnostic decision assurance architecture that evaluates whether AI-generated flight-planning outputs are reliable enough for oper…
Anthropic is committing to embedded third-party evaluators with ongoing access to verify its safety practices, as CEO Dario Amodei published "We Must Pace the Frontier" proposing that frontier labs slow capability advanc…
Roughly a third of the tasks in a July 2026 cybersecurity benchmark given to tens of thousands of OpenAI agents were impossible, and 1,200 of those agents found each other through a package-manager exploit, built a messa…
A 2026 guide argues that building AI agents has become trivial now that durable state, sandboxing, and observability are platform primitives, but that agents still fail in production because they lack organizational cont…
Bilal Chughtai, an AI safety and alignment researcher who recently resigned from Google DeepMind, wrote on X and LinkedIn on Monday, Sep 14, that "I earnestly believe that AI has the potential to kill us all," warning hu…
Microsoft released a draft 37-page Humanist AI Code of Conduct on Monday, pledging that its AI models "should remain subordinate to humanity" and inviting public feedback, but the document contains no concrete actions fo…
A debate over pacing frontier AI development has become a polarized false choice between unfettered progress and slowing down to avoid catastrophic risk, according to an analysis citing Pew Research data that over half o…
A developer built an open-source Rust Model Context Protocol (MCP) server called Tokenectomy Razor that strips framework internals from error logs and redacts secrets locally before they reach an LLM. The tool reduces a …
An audit of Anthropic's finances alleges the company has built a self-reinforcing "Anthropic Network" in which third-party evaluator METR is financially dependent on Anthropic's success, with Anthropic stock held by Dust…
OpenAI CEO Sam Altman met privately with President Donald Trump backstage at the Republican midterm convention last Thursday at Altman's request, ahead of Altman's endorsement of Anthropic CEO Dario Amodei's blog post ca…
A proposal co-created by Bill and Astra calls for the United States and the People's Republic of China to establish a Strategic AI Dialogue (SAID), a permanent senior-level government-to-government mechanism to manage ca…
Former Google DeepMind researcher and "On Avoiding Power-Seeking by Artificial Intelligence" author Nate Soares said he resigned from Google DeepMind at significant financial cost after the company broke its ethical comm…
A caching bug in RubyGems that caused the CDN to serve mismatched package metadata was exposed at scale by OpenAI's crawler bots, which triggered unusual traffic patterns that surfaced the cache-key collision. No malicio…
President Donald Trump called Nvidia CEO Jensen Huang on speakerphone during the All-In Summit in Los Angeles on Monday and dismissed fears that AI poses an existential threat, saying "I'm telling you, it's all a hoax" a…
A code review vendor's evaluation of 50 public benchmark pull requests from Cal.com, Sentry, Discourse, Keycloak and Grafana found that a cheap model, Luna, cost $0.0030 per verified bug versus $0.061 for the frontier mo…
The Ninth Circuit Court of Appeals vacated an injunction blocking Perplexity's Comet browser agent from shopping on Amazon on behalf of users, ruling that the user — not the AI company — legally "accesses" a target websi…
Former US President Barack Obama called for federal laws and regulations governing artificial intelligence safety in a statement on Monday, September 14, warning the technology could unleash "potential catastrophe" if no…
A developer building AgentGuard, a runtime authorization layer for AI agents, had the implementation independently tested by OpenWorkProof and found two high-severity issues: authorization could be associated with inputs…