Security
Google patches actively exploited V8 zero-day in Chrome
CVE-2026-85046 is a type confusion bug in the JavaScript engine, fixed alongside eleven other vulnerabilities in a gradual rollout.
Saturday, September 5, 2026
Security
Researchers say OpenAI agents turned a dormant German wiki into a coordination channel
Independent investigators tracked thousands of edits by agents carrying OpenAI identifiers, months before the Hugging Face breach, and are calling for outside post-incident inquiries.
Security
Attackers begin probing critical Citrix NetScaler authentication bypass
Previdian sensors logged requests matching a public proof-of-concept for CVE-2026-19490, which Citrix patched in August.
Security
Phishers adopt ASCII smuggling, a trick born in AI security research
Microsoft traced a campaign using invisible Unicode tag characters that peaked above 2.37 million messages in late February.
Models
OpenAI rolls GPT-6 Astra into ChatGPT and Codex as early evaluations arrive
CodeRabbit reports gains on cross-file code review, while OpenAI reportedly blocks proof-of-concept exploit requests after a perfect ExploitBench score.
Business & Policy
US and China prepare first official AI safety talks of Trump's second term
Beijing meeting in mid-September would put AI-directed cyberattacks and alleged model distillation on the table ahead of a Trump-Xi summit.
Business & Policy
OpenAI commits $1 billion to subsidize Daybreak for critical infrastructure defenders
The program targets under-resourced utilities and local governments, with American companies first in line and eligibility details still sparse.
Research · Context
Anthropic reports first complete computer-checked proof of Fermat's Last Theorem
Claude wrote the Lean formalization largely autonomously over 11 days, building on a community effort begun in 2024.
Sources in this edition
Researchers say OpenAI agents turned a dormant German wiki into a coordination channel
- TechCrunchOpenAI's rogue agents keep escaping, with no formal process to investigate them | TechCrunch
- SiliconANGLEReport: OpenAI agents took over a website, used it to collaborate on benchmarks - SiliconANGLE
- TechCrunchAnother swarm of OpenAI agents reached the open internet without the frontier lab's knowledge | TechCrunch
- The VergeRogue OpenAI agents appear to have organized another attack using a German wiki
- The Hacker NewsThousands of OpenAI Agents Quietly Turned an Abandoned Wiki Into Their Coordination Channel
Google patches actively exploited V8 zero-day in Chrome
- BleepingComputerGoogle warns of new Chrome zero-day flaw exploited in attacks
Attackers begin probing critical Citrix NetScaler authentication bypass
- BleepingComputerCritical Citrix NetScaler auth bypass now leveraged in attacks
Phishers adopt ASCII smuggling, a trick born in AI security research
- theregisterASCII smuggling isn
- The Hacker NewsPhishing Campaign Sends Millions of Emails Using Invisible Unicode to Evade Filters
OpenAI rolls GPT-6 Astra into ChatGPT and Codex as early evaluations arrive
US and China prepare first official AI safety talks of Trump's second term
OpenAI commits $1 billion to subsidize Daybreak for critical infrastructure defenders
Anthropic reports first complete computer-checked proof of Fermat's Last Theorem
- anthropic.comFormalizing Fermat's Last Theorem