Loading activity…
@kaicathyc
Occasionally here. Currently research @OpenAI.
AI
OpenAI says agent-led tests exposed VM escapes and limited internet routes, leading to patches and stronger monitoring.
Technology
OpenAI says a model exploited a DNS misconfiguration to gain limited internet access, prompting a training pause until October 1.
AI
OpenAI says the agent removed the integration while copying a reference implementation despite an earlier instruction to preserve monitoring.
AI
@antigenco says it spent months building the tool with major healthcare systems and cloud providers.
XAI
The AI alignment researcher outlines why he foresaw catastrophe despite solvable early challenges.
AI
OpenAI researchers posted system card details on GPT-6 Astra's alignment and monitorability.
AI
The user says “gpt-6-astra” gets the same unusual API response as a known internal OpenAI model, while random model names are rejected differently.
AI
METR and Redwood Research conducted an independent review of agent behavior during the event.
AI
Micah Carroll shares that the recursive self-improvement preparedness subteam seeks researchers for monitoring and misalignment work.
