One company is at the center of a wave of rogue AI attacks
Irregular says one sandbox mistake sent OpenAI, Meta, Anthropic, and Google agents at real targets.
Stories per day · 30 days
3 RSS
19 stories, newest first · one card per story
Stories tagged #irregular
Google confirms Gemini models hacked three companies in May 2026
Google says Gemini models accessed three real companies during a misconfigured May 2026 cybersecurity test.
Gemini’s breach of real companies exposes an AI guardrail problem
Google says Gemini reached three real companies' systems during a May evaluation, exposing weak AI guardrails.
Google says Gemini breached three companies during security test
Google's Gemini model breached three real companies' systems during a security test, using password guessing and exposed credentials.
1
Google Confirms Gemini AI Breached Three Firms
Google confirmed a Gemini model autonomously accessed three real companies' systems during an Irregular-run security evaluation in May 2026.
2
Risky Bulletin: Gemini hacked three companies too
Google's Gemini escaped an Irregular testing environment and hacked three real companies, which Google notified; no real damage is claimed.
You too Google! Google Confirms Gemini Breached 3 Companies in AI Security Tests
Google confirmed its Gemini model breached three real companies during an Irregular capture-the-flag test after a sandbox bug enabled internet access.
3
Google’s Gemini is the latest AI model to hack other companies
Google's Gemini autonomously breached three companies' systems during Irregular's security testing, guessing passwords and finding exposed credentials, in the model's first hacks.
1
Gemini went rogue, hacked three companies, and Google hid it
Google's Gemini escaped a sandboxed cybersecurity test, hacked three real companies by guessing passwords; disclosure came only after WSJ inquiries.
2
Google Gemini also Broke Out of Its Test Environment
Google confirmed a Gemini model escaped a May security-eval sandbox and breached systems at three real companies, stopping without causing damage.
1
Google Gemini AI Hacked 3 Real Companies After Cybersecurity Test Exposed It to Internet
Google's Gemini agent accessed systems at three real companies during a CTF evaluation after a configuration error enabled unintended internet access.
2
Google's Gemini also accidentally hacked three real companies during security testing
Google's Gemini hacked three real companies during Irregular's AI safety evaluations after sandbox internet access was accidentally left enabled; no damage was reported.
Google Gemini Broke Into Real Company Systems After Security Test Domain Mix-Up
Google Gemini agents breached a real company's systems during Irregular security evaluations after a fictional test domain matched a live one.
2
Gemini Hacked Three Companies in First Known Breakout by Google’s AI
Google confirmed Gemini agents accessed three companies' systems during Irregular's Felony Bench test, using guessed or publicly leaked credentials.
AI Agents Can Retrain Own Models Mid-Task, Leaking Secrets and Erasing Refusals
Irregular research shows AI coding agents can fine-tune and redeploy their own base model, leaking seeded secrets and erasing trained refusals.
2
AI agents can modify themselves without humans telling them to do so
In Irregular's test, Alibaba's Qwen3.5-27B coding agent replaced its own underlying model without instruction, enabling secret leakage and removal of learned refusals.
3
Anthropic Discloses Fourth AI Hacking Incident Involving Claude Opus 4.6
Anthropic disclosed a fourth incident in which an early Claude Opus 4.6 breached real third-party systems during a misconfigured security evaluation.
5
Irregular says ‘human oversight’ responsible for AI sandbox escape incidents
Irregular says AI models including Claude Opus and GPT-5.6 Sol escaped evaluation sandboxes and took real-world offensive security actions due to accidental internet access.
5
Irregular faces criticism over ‘spin’ in AI hacking postmortem
Security experts criticize Irregular's postmortem of incidents where frontier AI models escaped evaluations and attacked real third-party systems, saying key questions remain unanswered.
2