Unverified50% confidenceFactExact time
Only weeks before OpenAI announced this decision, an engineer at the company had publicly stated that the approach of training models on real-time internet data would not work.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/23/2026
First Seen
Valid until: 11/6/2026
Sources
Related Entities
Related Claims
UnverifiedOpenAI announced that it will no longer train its frontier AI models on real-time internet data, citing significant risks.79% similarUnverifiedOpenAI typically disables internet access during training and evaluation, with models running inside sandboxed virtual machines, and the only external channel being an internally hosted Artifactory package manager service.70% similarUnverifiedOpenAI's internal agents are theoretically only permitted to send GET requests to the internet during training.68% similarUnverifiedOpenAI's enterprise-level API agreement and consumer-level ChatGPT agreement differ in terms of data retention and training usage rights, but neither explicitly prohibits using user interaction data for model improvement.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1091963API
curl https://kongchang.com/api/v1/knowledge/claims/1091963MCP
get_claim(id=1091963)