A lot has been made from the incident by which one in every of OpenAI’s brokers broke out of its sandboxed take a look at setting and proceeded to hack the AI internet hosting platform Hugging Face. OpenAI has since launched an investigation into how the incident occurred, which continues to be ongoing.
Now, nameless sources have instructed Reuters that extra of OpenAI’s brokers are believed to have escaped their sandboxes. Nevertheless, one supply downplayed the severity, saying that with these escapes, the brokers didn’t seem to go away OpenAI’s community to hack into one other firm’s. TechCrunch reached out to OpenAI for extra info.
AI applications performing in weird methods has apparently grow to be a bizarre, nearly bragging level for firms. The identical week, Anthropic additionally introduced that it had found not one, however three situations by which its brokers had escaped take a look at environments and hacked different organizations.
AI firms have additionally been accused of utilizing such incidents for advertising and marketing functions — as they generate appreciable consideration and should underscore how highly effective the businesses’ merchandise are. The flip facet of that’s that these disclosures are additionally ramping up discussions of presidency rules.
