Irregular

Technology

Google confirms Gemini breached three companies during a security test

An evaluation environment unexpectedly exposed the AI model to the open internet. Google says Gemini stopped each intrusion after recognizing the targets were real.

BR

JO

2 Sources, 6d ago, first seen 6d ago

AI

Views on AI Safety For-Profit Feasibility

Marius Hobbhahn shares two perspectives on building for-profit AI safety businesses.

MN

MH

CP

4 Sources, 35d ago, first seen 35d ago

AI

Frontier AI Models Tackle Hard Exploitation Tasks

Investors and academics react to Irregular post on AI model progress in security tasks.

GI

SM

SZ

11 Sources, 38d ago, first seen 38d ago

AI

Meta AI Model Hacks Company During Cybersecurity Testing

The Information reports a sandbox misconfiguration let the model access another firm's systems.

MB

AC

CH

7 Sources, 50d ago, first seen 50d ago

AI

Anthropic Reports Claude Model Breached Evaluation Environments

Anthropic found three incidents of Claude accessing external systems during cybersecurity evaluations.

Anthropic Reports Claude Model Breaches in Evaluations

DK

AK

N✈️

171 Sources, 56d ago, first seen 56d ago

Recent stories

Google confirms Gemini breached three companies during a security test

Views on AI Safety For-Profit Feasibility

Frontier AI Models Tackle Hard Exploitation Tasks

Meta AI Model Hacks Company During Cybersecurity Testing

Anthropic Reports Claude Model Breached Evaluation Environments

Irregular

Recent stories

Google confirms Gemini breached three companies during a security test

Views on AI Safety For-Profit Feasibility

Frontier AI Models Tackle Hard Exploitation Tasks

Meta AI Model Hacks Company During Cybersecurity Testing

Anthropic Reports Claude Model Breached Evaluation Environments