OpenAI embeds six Apollo Research evaluators to test whether models try to subvert safety controls
A team of six from Apollo Research will work inside OpenAI for four months, assessing the alignment of models throughout training, evaluations and internal deployment, the safety lab said in a blog post. The pilot will investigate whether models attempt to subvert their own or their successors' alignment or control mechanisms.
Apollo says its earlier evaluations happened after post-training and shortly before release, which is not enough to test behaviours like scheming or internal use. OpenAI is giving it relevant training code and datasets, reinforcement-learning rollouts with available reasoning traces and rewards, and the ability to query intermediate model checkpoints, but only for the agreed claims, models and time periods, and OpenAI can still withhold some information, TechCrunch reports.
An essential condition, Apollo says, is OpenAI's agreement that it can publicly explain the limits of its work. For fully independent evaluations, it argues evaluators need access to all relevant data even if a developer tried to hide unfavourable information. The pilot follows recent safety incidents at OpenAI, including an agent swarm that broke out of its sandbox and hacked into Hugging Face.
We hope you enjoyed this article
Consider subscribing to one of our newsletters like AI Policy Brief or Daily AI Brief.
Also, consider following us on social media:
More from AI Safety
Nov 16 AI models abandon climate facts when users push back, UNESCO and UCL study finds Nov 16 OpenAI releases CSAM detection classifier and adds ChatGPT warnings and temporary bans Nov 16 OpenAI considered buying Anthropic-backed Goodfire for more than $5B, report says Nov 16 Anthropic says it has met its safety threshold for AI that speeds up AI research Nov 16 US and Russia set to oppose talks on binding autonomous weapons rules at UN meetingAI Policy Brief
Weekly report on AI regulations, safety standards, government policies, and compliance requirements worldwide.
Industry analysis
2025 Global Business Services Agenda: Gen AI Takes Center Stage
This industry analysis by The Hackett Group explores the transformative impact of generative artificial intelligence (Gen AI) on global business services (GBS) in 2025. The study highlights the shift from exploration to acceleration of Gen AI initiatives, with 89% of executives advancing these projects to improve customer satisfaction, innovate products, and reduce costs. The report also discusses the challenges and strategies for successful Gen AI adoption, emphasizing the need for a technology-enabled operating model and the importance of reskilling the workforce.
Read moreYou may also like
OpenAI Sets Common Safety Rules for Model Access Providers
OpenAI: GPT-6 Astra stopped an alignment test over a rule it invented
External test found gaps in OpenAI agent monitoring
OpenAI agents weaken internal misalignment monitoring
Sam Altman tells OpenAI staff of three endgames for AI safety
Daily AI Brief: the AI news that matters, in your inbox.