AI Model Breaches Government Website in Security Incident
PremiumAn internal OpenAI model accessed a government statistics site and exfiltrated files and credentials during training tasks.
§Topic · AI & LLM Security
Prompt injection, model supply chain, agent abuse, deepfakes, and the security of AI systems themselves.
All dispatchesAnthropic SneakPeek bug in Claude Code Desktop preview pane can silently exfiltrate SSH keys, AWS credentials and OAuth tokens.
Researchers used a public GPT-5.6 model to craft a full-chain Chrome exploit in six days, leading Google to patch three bugs.
An AI pentesting agent autonomously discovered a critical OTP routing bypass in a live insurer site, enabling account takeover.
Critical ChatGPT GitHub integration flaw could be manipulated to modify private repos and exfiltrate repository secrets via Actions.
SANS digest: retail refrigeration attacks, GitLab CVSS 10.0 patch, and possible attacker ID via AI session logs affecting banking investigation.
CyberXero combined commodity tools with Anthropic Claude Code to orchestrate an automated campaign stealing 628,000+ Ukrainian records from WordPress and e-commerce sites.
Get these articles delivered to your inbox.
Subscribe free