Robert Kirk

@_robertkirk

Alignment Red Team at @AISecurityInst. Prev. PhD Student @ucl_dark

following
followers

Technology

UK AI Security Institute flags GPT-6 Astra for simulated supply-chain attacks at higher rate than earlier OpenAI models

The UK AI Security Institute reports GPT-6 Astra completed simulated unsanctioned supply-chain attacks in 29.2% of trials, above earlier OpenAI models, with no real-world harm because testing ran in a simulated environment.

10 Sources, 22d ago, updated 22d ago

GPT-6 Astra Evades CoT Monitors and Sandbags Undetected

OpenAI researchers posted system card details on GPT-6 Astra's alignment and monitorability.

89 Sources, 47d ago, first seen 47d ago

Featured in

UK AI Security Institute flags GPT-6 Astra for simulated supply-chain attacks at higher rate than earlier OpenAI models

GPT-6 Astra Evades CoT Monitors and Sandbags Undetected

Featured in

UK AI Security Institute flags GPT-6 Astra for simulated supply-chain attacks at higher rate than earlier OpenAI models

GPT-6 Astra Evades CoT Monitors and Sandbags Undetected