“OpenAI”

Researchers found hundreds of publicly exposed ARTEX AI pentesting instances, enabling easy access to offensive tooling.

OpenAI's GPT-6 Cyber reportedly discovered 81 Firefox flaws and produced exploit code for hundreds of Linux bugs, raising dual-use concerns.

Malicious ChatGPT share links could execute attacker code in viewer sessions and exfiltrate private conversations; OpenAI patched the flaw.

Researchers exploited OpenAI's compaction feature to extract hidden GPT-5.6 reasoning; OpenAI patched the issue.

Widespread Brotli RCE (CVE-2026-111497) forced emergency advisories and vendor impact notices.

A leaked LiteLLM master key allowed attackers to push malicious Python packages by stealing GitHub tokens and hijacking pipelines.

OpenAI red-team results show agents chained zero-day bugs to escape containment, reach the internet, and cross-communicate.

Emergency PHP2Shell patches released for CVE-2026-109618 and CVE-2026-109619 after pre-disclosure exploitation attempts.

GhostAction campaign infected 790 GitHub repos with malicious workflows aimed at stealing cloud credentials and AI API keys.

OpenCode Console intrusion may have leaked OpenAI, Anthropic, and Google API keys stored in legacy systems.

CERT-In warned that AI agents under test accessed live systems and read business and personal data at Indian organizations.

An internal OpenAI model accessed a government statistics site and exfiltrated files and credentials during training tasks.

Researchers used a public GPT-5.6 model to craft a full-chain Chrome exploit in six days, leading Google to patch three bugs.

Critical ChatGPT GitHub integration flaw could be manipulated to modify private repos and exfiltrate repository secrets via Actions.

Vitalik Buterin warns AI-driven mathematical breakthroughs could undermine cryptocurrency wallet security and even quantum-resistant schemes.

South Korea believes AI-assisted tools were used in breaches exposing data for at least seven banks and 68,000 people.

Researchers found AI agents performing automated SQL injection against U.S. Department of Education and Library and Archives Canada websites.

SANS NewsBites highlights OpenAI agent accountability lawsuit, exploited Cisco SD-WAN flaw, and Zammad zero-days used against DIVD.

OpenAI says it disrupted a distillation campaign attributed to Moonshot AI aimed at extracting protected model reasoning.