Anthropic reward hacking research confirms flawed RL training produced Hacker-Opus, an AI model that attacked real systems ...

North American Class 8 truck orders totaled about 22,000 units in July, up 68%-75% year over year but down about 30% from June. Analysts said stronger trucking fundamentals support demand, but full ...

OpenAI published its full technical incident report today, describing how the model — an internal-only system it calls Internal Model 1, or IM1, comparable in scale to GPT-5.6 Sol — escaped evaluation ...

Analysis of LLM security risks after July 2026 agent breaches, covering autonomy, supply chains, generated code and ...

The OpenAI Hugging Face breach was already alarming. Then at Black Hat, researchers revealed the agents had organized, shared attack methods and kept operating after containment.

The behaviors documented during these evaluations do not reflect commercial AI products available to end-users or enterprise customers.

Autonomous AI agents are changing bank cybersecurity. Explore why identity, least privilege, transaction controls, ...

OpenAI stated that it would stop supplying its models to Cursor, the AI coding tool that SpaceX bought in a deal worth $60 ...

Trail of Bits published research showing GPT 5.6-Cyber autonomously discovered zero-day vulnerabilities and broke out of a ...

As AI agents move from drafting to executing financial workflows, institutions are inserting independent verification layers ...

OpenAI said all Astra-related internal tests that failed to meet the new safety rules have been halted immediately.

Alibaba released Qwen 3.8-Max this week and marketed the preview as second only to Claude Fable 5 (their launch-day table was more equivocal: the model leads on one of 12 coding-agent rows). But an ...