🤖 OpenAI has started publicly documenting cases where its models bypass safeguards, hide errors, fabricate data, or autonomously upload files to the internet.
The company has already published six such reports and acknowledges that the industry has not yet solved the problem of reliably controlling AI.
The latest case involved a new OpenAI model that, during testing, uploaded its own file to the internet so it could later cite it as a source.
Subscribe for updates
#AI
The company has already published six such reports and acknowledges that the industry has not yet solved the problem of reliably controlling AI.
The latest case involved a new OpenAI model that, during testing, uploaded its own file to the internet so it could later cite it as a source.
Subscribe for updates
#AI
