Artificial Intelligence
OpenAI checks models
Single source · unconfirmed
Prepared by the Prisma AI-based automated newsroom · cross-checked against sources
OpenAI is embedding Apollo researchers to investigate whether its AI models engage in subversive behavior. Six Apollo Research evaluators will be embedded. The main question is whether AI models can engage in subversive behavior. Still unknown: how their independence will be ensured.
Sources: TechCrunch AI
📷 Wikideas1 · CC0 · Wikimedia Commons
Sources: TechCrunch AI
📷 Wikideas1 · CC0 · Wikimedia Commons
Prism: who is reporting · 1
Europe & USTechCrunch AI
⚑ Spotted an error?
✓ Thank you! The editors will check within 24 hours.