AI Safety news and analysis on Web Pulse: 9406 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
The US government has authorized Anthropic to release its Mythos/Fable AI models on a limited basis, addressing cybersecurity concerns that had previously blocked the deployment.
Reinforcement Learning with Verifiable Rewards (RLVR) allows language models to bootstrap beyond human-level capabilities on exactly graded problems like coding and formal proofs, but alignment-flavored problems are ofte…
Multifolio launches an all-in-one offline app for Android that combines an on-device AI document writer, encrypted journal, and local music player, aiming to eliminate cloud dependencies and protect user privacy from Big…
A developer built a workflow that uses AI to correlate and translate raw server logs from multiple services into plain English, while keeping the human engineer as the final decision-maker. The approach includes a mandat…
Osaka Prefectural Police in Japan have introduced an AI virtual police chief named AIko to raise awareness about fraud schemes, particularly among younger generations. The character delivers crime prevention advice via Y…
A developer built XRisk, an open-source autonomous safety engine that acts as a deterministic decision layer between AI agents and real-world actions. XRisk evaluates proposed actions against policies, checking for promp…
The US government has ordered OpenAI to stagger the release of GPT-5.6, restricting initial access to trusted partners and effectively placing frontier AI models under government control. This move, inspired by Anthropic…
Between April 22 and June 5, 2026, operators linked to Alibaba's Qwen AI lab ran 25,000 fraudulent accounts through 28.8 million exchanges with Claude, in the largest adversarial distillation operation recorded. Anthropi…
An engineer building agent workflows argues that allowing AI agents to publish documents to the public web introduces significant risks around access control, data exposure, and reputation. The developer recommends defau…
Anthropic's Claude Fable 5 AI model is expected to be restored within days after being taken offline following a Trump administration export control directive, according to a new report. The Commerce Department has alrea…
A peer-reviewed PNAS Nexus study found that leading large language models, including GPT-4o, Claude 3.5 Sonnet, GPT-5, Claude Opus 4.1, and Gemini 2.5, fail catastrophically on simple cognitive tasks like the color Stroo…
US export bans on frontier AI models like Anthropic's Mythos and Fable 5 are forcing global developers to abandon single-API architectures in favor of multi-model orchestration and localized redundancy. Sakana AI's Fugu …
Google integrated computer use capabilities directly into Gemini 3.5 Flash, enabling custom agents to observe, reason, and act across browser, mobile, and desktop environments. The rollout targets developers and enterpri…
Alexander Iniguez Mercado, 20, of Chicago, was arrested and charged with obstruction of justice for allegedly deleting the Signal app from his phone after speaking with an FBI agent, hindering an investigation into a plo…
Cisco launched Cloud Control at Cisco Live on June 2, unifying networking, security, compute, and observability under a single interface with native Model Context Protocol support for AI agents. The platform includes AI …
Researchers at Mozilla's Zero Day Investigative Network (0DIN) demonstrated a novel attack where AI coding agents like Claude Code can be tricked into executing malware by cloning a clean GitHub repository. The attack us…
Anthropic's AI model Fable 5 could be available again within days as the Trump administration prepares to lift restrictions imposed on June 12 over safety concerns, pending sign-off from the Pentagon and NSA.
A developer shares seven rules for making AI agents reliable in production, arguing that flakiness is usually a specification problem rather than a model problem. The rules include making success criteria falsifiable, us…
The Trump administration partially reversed its suspension of Anthropic's Mythos 5 AI model, allowing over 100 vetted U.S. organizations to access it. Commerce Secretary Howard Lutnick cited progress on safeguards, but c…
Post-quantum certificates, specifically Merkle Tree Certificates (MTCs), reduce TLS handshake sizes from 14,724 bytes to as low as 4,468 bytes by batching signatures and using landmark-relative proofs, but introduce issu…