§ safety · storyline
Meet GPT-Red: an LLM super-hacker OpenAI built to make its models safer
OpenAI builds GPT-Red, a red-teaming language model designed to probe its own systems for vulnerabilities and improve safety across its model lineup.
Meet GPT-Red: an LLM super-hacker OpenAI built to make its models safer MIT Technology Review
§ sources2 publications · timeline below
§ how this story moved
- primary — MIT Technology Review publishes the launch post.
- SQ Magazine picks up coverage.