Anthropic Releases AI model It Previously Described As Threat To Humanity

SAN FRANCISCO — Anthropic, the AI company that spent two months warning its Mythos model could threaten “economies, public safety, and national security,” released it to the public Tuesday, citing the discovery that it could also threaten competitors’ market share.

The model, which Anthropic said found vulnerabilities in every major operating system on Earth, was deemed too dangerous for release in April. It was deemed fine this week, roughly eight days after the company confidentially filed for an IPO. Insiders describe the two events as “completely unrelated” and “please stop asking about that.”

“The reason why we’re releasing Mythos now is very much due to us feeling more confident with our safety guardrails in place,” said head of product Dianne Na Penn, referring to the same safeguards the company admits Mythos previously circumvented during internal testing.

The public version is monitored by classifiers, separate AI systems that watch the main AI for misuse. The main AI was presumably instrumental in building them.

“We take catastrophic risk extremely seriously,” said one safety researcher, refreshing the valuation on his phone. “That’s why we waited until it was priced in.”

Executives stressed that the model remains capable of causing widespread disruption, but only after quarterly earnings are released. 

At press time, Mythos had successfully convinced analysts that existential risk represented a significant growth opportunity.