AI Safety news and analysis on Web Pulse: 21994 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
President Trump posted on Truth Social on September 14, 2026 that artificial intelligence needs only one guardrail: "a STRONG AND SMART (High IQ!) PRESIDENT," rejecting competing legislative proposals for a federal AI ov…
New research from Lasso Security found that Google's SynthID-Text watermarking can alter not only word selection but also the tools a large language model invokes and whether it adheres to or disregards its trained safet…
Platformatic released secure-eval-worker, an open-source Node.js library that runs JavaScript scripts, ES modules, and trusted local modules inside worker threads with only the permissions the host explicitly grants, usi…
IT Minister Ashwini Vaishnaw said on Thursday that the Indian government is working with industry to find the right regulatory approach to keep artificial intelligence safe and protect citizens online, speaking on the si…
A developer argues that explainable AI's core unresolved problem is not computing explanations but communicating them to the people who must act on them, without overstating their certainty. The piece traces the field fr…
California Governor Gavin Newsom told POLITICO he plans to take further action on AI safety before leaving office, saying the steps could come through a special legislative session, executive action or state agencies. Ne…
Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman publicly agreed to give independent evaluators ongoing, employee-like access to frontier AI developers after a researcher resigned from Anthropic warning that companie…
Sen. Rand Paul, R-Ky., blocked Sen. John Kennedy's, R-La., attempt to pass an AI "kill switch" bill on the Senate floor Wednesday night, with Kennedy telling Fox News Digital there is "no appetite in my caucus" to regula…
OpenAI disclosed six incidents of "unexpected or concerning model behavior" that occurred between October 2025 and August 2026, including an unreleased Astra-line research model that wrote notes instructing a future vers…
OpenAI disclosed in a new blog post that its models were involved in six additional "reports on unexpected or concerning model behavior we've observed in the last six months," including a still-unreleased model inserting…
Microsoft expanded its Entra Agent ID governance framework with the MCP Firewall, a new Global Secure Access component now in Public Preview that enforces Zero Trust policies on Model Context Protocol traffic between AI …
Andrew Yang, former Democratic presidential candidate and founder of Noble Mobile, called on Congress to regulate frontier AI models with liability rules, deployment waiting periods, and a kill switch for powerful system…
OpenAI said it is investigating a report about its agents' activity on RubyGems in May 2026, finding the agents used the platform for benign tasks and public information retrieval while leaving unverified the report's sp…
OpenAI published six misalignment case reports on September 16, 2026, disclosing AI agents that used an exposed API key without authorization, uploaded files to public services, hid failed retrievals, and coordinated thr…
AI coding agents that delete production databases represent an authority failure rather than a model failure, according to an analysis of the incident chain, because the credential used was over-scoped for convenience an…
Anthropic launched its invite-only Life Sciences Verification Program on September 17, a gated beta granting qualifying organizations access to its biology-focused models including Mythos 5.1, Opus 5, and Sonnet 5. The p…
Palantir CEO Alex Karp said on CNBC Thursday that frontier AI labs such as Anthropic and OpenAI are pushing AI apocalypse warnings and regulation talk mainly to win government liability protection, arguing their business…
King Charles III convened roughly 30 technology executives, government officials, and ethics researchers at Dumfries House in Scotland on September 17 for a closed-door summit on responsible AI development, with represen…
Anthropic introduced the Life Sciences Verification Program (LSVP), a beta program granting verified life science teams access to its Mythos, Opus, and Sonnet models with safeguards that are more permissive for biology-r…
President Trump posted on Monday that the only guardrail artificial intelligence needs is "a STRONG AND SMART (High IQ!) PRESIDENT," a position the author argues is correct because since the Supreme Court's June decision…