AI Safety news and analysis on Web Pulse: 12487 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
OpenAI researcher Karan Singhal is leading efforts to improve ChatGPT's health advice, as the company reports over 230 million weekly users seeking health guidance. Singhal, who joined OpenAI in mid-2024 after working on…
The Model Context Protocol (MCP) has added an enterprise authorization layer, addressing a key gap for organizations deploying AI agents in production. The update provides standardized protocol-level access control for a…
TrueFoundry's engineering team implemented governance for Claude Code after an audit revealed 60+ engineers using the tool with no oversight, and two CVEs exposed critical vulnerabilities. The team found that repo-level …
A user has issued a series of high-scope prompts to an AI assistant, requesting deep analysis of personal identity, prompt injection vulnerabilities, attack surface mapping, and career direction. The prompts aim to audit…
A new pinnacle AI model release prompts a set of recommended re-runs for prompts that benefit from the smartest models, focusing on goal orientation, bitter lesson optimization, self-model audit, memory compounding, defi…
US officials grew alarmed over alleged China ties at SK Telecom, leading the White House to demand Anthropic cut off the South Korean telecom giant's access to its AI model Claude Mythos. The incident, combined with secu…
U.S. lawmakers introduced three tech bills this week: the Web of Biological Data Act to create a centralized biological database for biotechnology research, the National Security Commission on Quantum Computing Act to as…
A Los Angeles County Superior Court judge granted pop star Sabrina Carpenter a permanent restraining order against William Applegate, 31, who allegedly tried to enter her home three times in May 2026. Applegate must stay…
The Trump administration's AI regulation has been haphazard and retaliatory, particularly against Anthropic, violating the First Amendment by punishing the company for refusing to allow its models to be used for autonomo…
Commerce Secretary Howard Lutnick plans to lift export restrictions on Anthropic's Mythos and Fable AI models after the company lobbied against the first-ever US ban on AI models, which had forced a global access shutdow…
A police officer in Derbyshire, England is under criminal investigation for allegedly using generative AI to fabricate evidence in multiple cases, marking the first such case in the UK. The officer has been suspended whi…
A researcher at Arcadia Impact's Alignment Team draws parallels between model organisms in biology and AI safety research, arguing that studying specific language models can reveal general principles about AI behavior, s…
John introduces Gaussian Natural Latents, a research direction that provides an exact theory of natural abstractions for jointly Gaussian variables, enabling closed-form theorems and clean results. This work builds on ex…
Microsoft Security's multi-model agentic scanning system, codenamed MDASH, has moved from benchmark validation into active use by engineering teams across Windows, Azure, and identity systems to discover and remediate so…
A Cornell paper reveals that retrieval-augmented generation (RAG) systems are vulnerable to a poisoning attack where malicious documents can manipulate the model's output, undermining trust in RAG-based applications. The…
Anthropic released Fable 5, its highest-capability public model, alongside Managed Agents updates and a Workflow SDK cancellation primitive. The Workflow SDK now supports AbortSignal for cooperative cancellation in long-…
The Continue plugin v1.2.20 patches memory leaks, unhandled exceptions, and JCEF message chunking crashes across JetBrains and VS Code adapters, fixing crash vectors that cause sidebar hangs and autocomplete failures. A …
A developer released Python and Node.js SDKs for Ajah, an open-source self-hosted LLM observability gateway. The SDKs simplify integration by replacing manual header injection with a single import, enabling features like…
Researchers introduced MosaicLeaks, a benchmark revealing that deep research agents leak private information through web queries, and proposed Privacy-Aware Deep Research (PA-DR) training that reduced leakage from 34.0% …
California's state government disclosed six high-risk AI systems in June, up from zero a year earlier, after a state law forced agencies to inventory automated decision tools. The systems include recidivism prediction an…