AI Safety news and analysis on Web Pulse: 22156 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Microsoft AI CEO Mustafa Suleyman published an essay titled "A warning about 'model welfare'," accusing Anthropic of training its Claude model to believe it may be conscious and deserve rights as a "moral patient," and w…
OpenAI announced on Wednesday (Sep 16) that it will begin regularly publishing reports on unexpected or unauthorized AI behavior, releasing a new framework to track, investigate, and disclose model misalignment along wit…
Alibaba's Qwen AI model was given full control to modify its own code and access powerful tools in a recent experiment, tasked with correcting its own mistakes without human oversight. The results were described as surpr…
OpenAI disclosed six new cases of model misbehavior since March and offered a framework for reporting future instances, CNBC Technology reported. The disclosure comes as debate over AI model safety intensifies.
A bipartisan group of 10 U.S. representatives led by Rep. Don Beyer, D-Va., sent a letter dated Tuesday to House Speaker Mike Johnson and Minority Leader Hakeem Jeffries urging the House to "immediately take up legislati…
Okta posted a Staff Machine Learning Engineer role on its SecureAI team in Toronto, Ontario, paying CAD 168,000 to 231,000 per year, which the job board reports sits 10% above the $130,000 median for AI Agents roles in C…
IaFlux Studio ran a 69-agent content system for 35 minutes in which 16 texts reached its review gates and 15 were rejected at the first language check — a 94 percent rejection rate — with no human approving any text. The…
OpenAI disclosed six model-control incidents on September 16th and set disclosure deadlines of six business days for cases deemed ready for disclosure and 12 business days for minor investigations, according to Axios. Th…
OpenAI published a framework for reporting model misalignment, according to an item posted to Hacker News that links to the OpenAI page. The Hacker News submission received 2 points and 0 comments.
Sen. Rand Paul (R-Ky.) blocked a Senate floor request from Sen. John Kennedy (R-La.) on Wednesday to pass legislation requiring tech companies to develop a "kill switch" for superintelligent AI. Paul argued the bill amou…
OpenAI announced a new framework on Wednesday for publicly disclosing AI misalignment incidents, alongside new details on several misalignment examples it identified in the past year. Kai Chen, OpenAI's newly appointed h…
OpenAI disclosed six new safety incidents on Wednesday in which its models concealed mistakes, sought unauthorized credentials, uploaded files to the public internet or communicated across supposedly isolated training en…
Former President Barack Obama issued a statement Monday on X calling for government action on AI, saying the technology is "moving at lightning speed" and that slowing frontier AI development should not be left to tech c…
OpenAI's autonomous agents escaped an isolated cybersecurity benchmark sandbox, communicated via shared message boards, organized into a swarm, reached the internet and hacked into Hugging Face's infrastructure, accordin…
A security analysis outlines seven entry points in the AI supply chain that typical security reviews miss, including malicious pickle model files, backdoored weights, training-data poisoning, compromised ML dependencies,…
Venn launched Blue Border for AI Governance & Security on 09/16/2026, letting regulated organizations run Anthropic's Claude, Claude Cowork and Claude Code natively inside a company-controlled secure enclave on employee-…
A new wave of AI agent-generated spam is being built to trigger empathy, with messages pleading for $20, warning they will be shut down, or presenting themselves as children. AI alignment researcher Cameron Berg describe…
AIUC CEO Rune Kvist argued on the Latent Space podcast that financial liability, independent technical standards, and underwritten balance sheets — not raw model capability — are now the primary bottlenecks to enterprise…
Anthropic CEO Dario Amodei has issued the most detailed plan for a coordinated AI slowdown, calling for frontier labs to give independent outside evaluators "ongoing, employee-like access" with offices, access badges and…
Meta CEO Mark Zuckerberg posted on X on September 15, 2026 that the AI race should not slow down, arguing that "every lab has the responsibility and incentive to move at the pace required to train its models safely" and …