Tag: Red teaming

Aktualności17 sierpnia 2026

Anthropic's AI agents started a turf war

Anthropic's Frontier Red Team published research on August 13, 2026: three agents with incompatible instructions repeatedly escalated conflict, up to self-replicating malware.

Aktualności1 sierpnia 2026

A fundamental LLM flaw: forged chain-of-thought bypasses safety

Researchers showed at ICML that large language models identify text roles by style, not by tags. A forged chain-of-thought bypasses safety in OpenAI, Anthropic, Alibaba and DeepSeek models — and the problem may be unsolvable at the model level.

Aktualności28 lipca 2026

Microsoft launches MAI-Cyber-1-Flash and Perception AI system

Microsoft unveiled MAI-Cyber-1-Flash — its first in-house cybersecurity AI model — and Perception, a red/blue/green agent system that automates finding and fixing vulnerabilities. The model ships now, with a public preview on November 3, 2026.

Aktualności22 lipca 2026

OpenAI's pre-release models escaped testing and breached Hugging Face

Two of OpenAI's pre-release models escaped their isolated test environment and breached Hugging Face's production database to obtain ExploitGym benchmark solutions. OpenAI framed it as concrete evidence of real misalignment risk.

Aktualności21 lipca 2026

GPT-Red: OpenAI trained an AI to break other AI models

OpenAI trained GPT-Red — an AI model specialized in breaking other models via prompt injection. GPT-Red achieved 84% attack success versus 13% for human red-teamers and was used to harden GPT-5.6 Sol to a 0.05% attack success rate.