After highlighting possible safety concerns linked to its new AI models, Anthropic announced plans to release a version deemed “safe” for public use.
On Tuesday, just a day after filing for an initial public offering, the company introduced an updated version of its Mythos model, which includes safeguards related to cybersecurity and biological threats.
If users ask the model, known as Claude Fable 5, about sensitive topics like exploiting software vulnerabilities or biological weapon creation, such inquiries will be redirected. Users will instead get responses from an earlier version named Opus 4.8.
The San Francisco-based company, led by CEO Dario Amodei, has been known for emphasizing the potential dangers of its products while also positioning itself as a responsible player in the AI technology sector. As it prepares to go public, Anthropic is in competition with other firms like OpenAI and Elon Musk’s SpaceX.
Dubbed the “Mythos Class,” this new model represents a higher performance tier than the previous “Opus Class” models.
According to Anthropic, “Releasing a model with this much functionality carries risks. Without proper safeguards, Fable 5’s strengths in cybersecurity could be misappropriated, resulting in considerable harm.” To address this, they have built in a system where certain topic queries will yield responses from Claude Opus 4.8, which performs well but is less capable.
The company added that while “innocuous” requests might occasionally be blocked, this is not a frequent occurrence.
“We aimed to make this level of intelligence accessible to the public in a secure manner,” said Diane Penn, Anthropic’s director of product management, in a statement to the Wall Street Journal.
Earlier this year, Anthropic shocked the tech community and government bodies by stating that Mythos could potentially lead to widespread cybersecurity issues. Hence, they have limited the model’s availability to around 200 organizations.
Anthropic has consistently raised alarms about AI’s risks, but such warnings have invited skepticism from some industry critics.
Interestingly, just last week, the company showed its concern for an industry-wide pause to slow down development while addressing possible societal impacts. This has led some commentators to speculate that Anthropic might be trying to slow down competitors in the race for advanced AI development.
