In April, Anthropic unveiled a new artificial intelligence system called Claude Mythos.
But the company said it couldn’t share its new A.I. with the public because Mythos could become a powerful tool for hackers looking for ways to break into computer networks. That warning caused panic among executives and government officials who worried that the technology was about to usher in a new era of cybersecurity threats.
Now Anthropic is releasing a straitjacketed version of Mythos that researchers believe is safe for widespread use and relies on older technology for some of what it does, the company said on Tuesday.
Called Claude Fable 5, this new system includes additional guardrails designed to block responses related to cybersecurity, biology and other vulnerable areas. Because of these guardrails, hackers may struggle to attack computer networks using Fable. But businesses and cybersecurity experts may also struggle to defend networks using the new system.
Most queries from Claude users in areas that could be perceived as too risky, the company said, will be handled by Claude Opus 4.8, which was released last month and was also designed to avoid the security risks of Mythos.
“That is the core of why we have been able to release Fable,” said Anthropic’s head of product management, Dianne Penn.
Anthropic shared the initial version of Mythos with about 40 organizations that maintain critical computer infrastructure so they could use the system to patch security vulnerabilities before hackers exploited them. This month, Anthropic said it would share Mythos with 150 additional organizations.
Cybersecurity experts disagreed on whether Anthropic was right to limit the release of the technology. Some applauded the company for restricting access to Mythos, while others criticized Anthropic for not sharing it with a wider pool of researchers who could try it, get a handle on what it can and cannot do and start using it for defense.
Mythos, which is still available to a limited number of organizations, does not have the guardrails built into Fable.
Outside researchers have shown that guardrails placed on A.I. systems are not always reliable, a point that Anthropic researchers recognize and say they are working on. “This is why we have invested so much in safety research,” Ms. Penn said. “It is a continuing progression.”
Fable tops all other publicly available A.I. technologies in overall performance, according to benchmark tests from Vals AI, a company that tracks the performance of the latest A.I. technologies. Its overall performance is 5 percent higher than Claude Opus 4.8, according to Vals AI’s tests.
The new system is also twice as expensive as Opus 4.8.
Fable is particularly adept at generating computer code and in mathematics, said Rayan Krishnan, the chief executive of Vals AI. But other systems still outperform Fable in other areas, including tasks involving health care and tax evaluation.
Over the past several months, companies in the United States, China and other parts of the world have built A.I. systems that are particularly good at writing computer code. If an A.I. system can write code, it can potentially identify vulnerabilities in software applications.
Hackers can then use the technology to attack computer networks. But tech companies, other businesses and cybersecurity experts can also use the technology to defend networks.
A week after Anthropic unveiled Mythos, the company’s chief rival, OpenAI, said it was sharing similar technology with a limited group of partners. But the company shared its model, GPT-5.4-Cyber, with a much larger group — hundreds of organizations. It said it would then release it to thousands more partners in the coming weeks.
(The New York Times sued OpenAI and Microsoft in 2023, claiming copyright infringement of news content related to A.I. systems. The two companies have denied those claims.)
The post Anthropic Releases ‘Safe’ Version of Its Mythos A.I. Technology appeared first on New York Times.
