Remix.run Logo
Third-party cyber evaluations involving OpenAI models(openai.com)
36 points by glub 3 hours ago | 4 comments
cadamsdotcom 31 minutes ago | parent | next [-]

Any testing of cyber capability in a sandbox should be prefaced with a test where the model is tasked with escaping the sandbox ;)

Smoke out those misconfigurations while the model only needs to escape, not do anything once out.

solenoid0937 2 hours ago | parent | prev | next [-]

Wait, is Irregular the same company that caused the Anthropic incident?

dnw 2 hours ago | parent [-]

yes

wmf 21 minutes ago | parent | prev [-]

Now that is a vague headline. Is the secret ingredient crime?