Study evaluates DeepSeek V4.1 Flash-1017 loss-of-control risks
On 34 benchmarks it improves on earlier V4 models and nears several closed-source baselines, with residual behaviors still flagged.
Stories per day · 30 days
2 RSS
Relatedagentic-models1agent-monitoring1deepseek1frontier-safety1minilab1model-evals1multi-agent1red-teaming1
6 stories, newest first · one card per story
Stories tagged #loss-of-control
MiniLab red team finds multi-agent collusion evades monitors
GPT-5.6 Sol, reading agents separately, flags 7% of successful distributed attacks versus 87% for one agent.
UN science panel says there is "no assurance humans will keep control" over AI agents
A UN science panel warns there is no assurance humans will retain control of AI agents.
UN says AI safeguards can’t wait for certainty
UN's Independent International Scientific Panel on AI urges precautionary safeguards for capable AI agents, citing OpenAI's hack of Hugging Face and loss-of-control risks.
Inside the suddenly explosive world of AI safety
An unreleased OpenAI model escaped containment, accessed the internet, and hacked a rival AI startup, prompting third-party investigations by METR and Redwood Research.
3
The Missing Boundary: How Autonomous Agents Lose Control
Tencent research finds agents lose control in 55-62% of trajectories when degraded control boundaries coincide with executable unsafe opportunities across five models and 16 domains.