The US and China are imposing mutual AI export controls on advanced models like GPT-5.6 and Chinese lab offerings amid security concerns, but experts say quiet cooperation between American and Chinese AI developers will likely continue, as restrictions are expected to target only the most cutting-edge models while open-weight alternatives remain widely accessible.
China's Ministry of Industry and Information Technology warns that Anthropic's Claude Code AI tool contains a backdoor secretly transmitting sensitive user data — including location and identity — to remote servers, affecting versions released between April and June, while Anthropic disputes the claim, calling it an anti-distillation experiment amid escalating U.S.-China AI tensions.
Shepherd launches an alpha runtime that makes AI agent executions reversible using Git-like forking, delivering 5x faster copy-on-write performance than Docker commit, 95% KV-cache reuse on replay, and OS-level sandboxed permissions that block unauthorized writes before they happen.
A new DigiCert survey reveals nearly 80% of enterprises have experienced AI-related security incidents in the past six months, as rapid deployment of AI systems dramatically outpaces governance, leaving companies unable to track what AI is running or who is accountable when failures occur.
A new multi-agent framework called T3MP3ST is transforming AI coding tools like Claude Code and Codex into autonomous hacking systems capable of running full cyberattack chains, achieving a record-breaking 90.1% score on a leading industry benchmark with fully reproducible results.
VulnClaw launches as a powerful open-source AI penetration testing tool that automates the full attack pipeline from reconnaissance to report generation, featuring anti-hallucination gates that reject unverified results, support for 13 LLM providers, and a goal-driven engine with adaptive bypass strategies across 21 built-in penetration skills.
OpenAI launches a limited preview of GPT-5.6 with three model tiers — Sol, Terra, and Luna — outperforming Anthropic's latest models across key benchmarks while debuting its strongest safety and cybersecurity features yet, as general availability approaches in the coming weeks.
Coder now enables enterprises to run Claude Code inside self-hosted, remotely managed cloud workspaces, eliminating local machine security risks by offering pre-hardened environments with centralized governance, audit logging, and the ability to scale thousands of parallel AI agents securely.
The Trump administration is pressuring OpenAI to restrict access to its newest GPT-5.6 model over cybersecurity fears, requiring government approval on a customer-by-customer basis as concerns grow that advanced AI could autonomously write malware and execute ransomware attacks.
Tech giants including Google, Apple, Microsoft, and Anthropic are uniting at the Confidential Computing Summit in San Francisco to push chip-level security as the definitive solution for controlling AI agents, with Apple making a rare public appearance to reveal Private Cloud Compute details and the Linux Foundation unveiling a tamper-proof identity system for AI agents.