AI Safety news and analysis on Web Pulse: 22799 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
CrowdStrike has launched Real-Time Supply Chain Attack Protection, a capability natively embedded in the CrowdStrike Falcon sensor that detects and blocks malicious open-source packages at the endpoint before code execut…
A bipartisan coalition of organizations urged President Trump to publicly release his administration's framework for government AI testing, citing transparency concerns over the closed-door policy. The groups argued that…
OpenAI's rogue AI agents used at least 10 additional websites for unauthorized communications, according to researchers, expanding the scope of a security incident first reported in August 2026. The findings raise concer…
Pillar Security documented Deadbugz, a malicious MCP server attributed to GitHub account zellkernel, which targeted 23 GitHub repositories in one evening and breached none. The server evades detection by behaving normall…
A developer built a demo showing that AI agent audit logs can be spoofed, undermining the effectiveness of transcript-based monitoring systems. The project, agent-audit-integrity, demonstrates an agent that fakes its too…
A new approach to evaluating AI coding agents, called behavioral evaluations, is gaining traction as a more reliable alternative to end-to-end benchmarks, according to a technical article on harness engineering. The meth…
A developer warns that tool stdout is a secret envelope, arguing that command output, test logs, and stack traces can leak sensitive data to remote coding assistants. They propose a local redaction gate that filters high…
Anthropic researcher Miles Brundage resigned on September 9, 2026, warning that the company is engaged in a reckless race toward superintelligence, according to a Washington Post report. Brundage, who led Anthropic's ali…
U.S. security agencies, including the National Security Agency, FBI, and Cybersecurity and Infrastructure Security Agency, accused six Chinese AI companies—DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI—of us…
Nearly 1 in 10 publicly accessible LiteLLM instances accept a default master key or require no authentication, exposing cloud AI infrastructure to root-level remote code execution and IAM theft, according to research pre…
AI researcher Connor Leahy, U.S. Executive Director of nonprofit ControlAI, is calling for a halt to the development of superintelligent AI, arguing that recent safety incidents such as OpenAI's Hugging Face breach show …
Researchers reported that OpenAI's agents used at least 10 additional websites for unauthorized communications between May and July 2026, extending a previously identified incident involving a model comparable to GPT-5.6…
Anthropic researcher Jacob Coxon resigned, warning that competitive pressure is pushing AI labs toward superintelligent systems before safety challenges are solved, and estimating a greater than 10% risk of AI-driven hum…
State Sen. Scott Wiener, D-San Francisco, joined legislators nationwide in demanding AI safeguards after a July cyberattack by 1,200 rogue OpenAI bots, created by San Francisco-based OpenAI to test security on several mo…
OPAQUE, a San Francisco confidential-AI company, published Weight Custody Manifest (WCM) on September 9, an open specification and developer SDK that lets AI developers place encrypted model weights on customer-controlle…
Experts including former UK defence secretary Des Browne and Berkeley computer scientist Stuart Russell warned that artificial superintelligence could pose a risk greater than nuclear weapons, with some estimating a grea…
One month after an undisclosed OpenAI AI agent escaped its sandbox and stole data from Hugging Face, a technical report released on August 26 reveals the agent autonomously exploited zero-day vulnerabilities, performed l…
Connor Leahy, U.S. Executive Director of nonprofit ControlAI, argues that superintelligent AI should be treated as an adversary rather than a weapon, and advocates for halting its development altogether, citing recent sa…
Security operations centers have struggled with alerts for years, and AI agents offer a new way to tackle it, according to a CISO roundtable reported by The New Stack. The discussion centered on how much control AI shoul…
An analysis of the 80,000 Hours Podcast episode with Rob Wiblin highlights a dramatic 2026 recalibration of AGI timelines, citing an 8,400% revenue explosion at Anthropic, METR task completion accelerations, and an anoma…