Artificial intelligence company Anthropic has suspended access to its cyber-focused AI models, Mythos 5 and Fable 5, after receiving a US government order requiring the company to block access for all non-US citizens, citing national security concerns.
The company said it received the directive on Friday and subsequently disabled access to both models for all users, including US citizens, while it works to comply with the order and assess its implications. Anthropic stated that it disagrees with the government's decision and believes the concerns may be based on a misunderstanding of the models' capabilities.
Government Concerns Focus on Security Safeguards
According to Anthropic, US authorities have been informed of methods that could potentially bypass some of Fable 5’s safety protections, a development the company believes prompted the government’s intervention.
Anthropic said officials provided details of a limited technique that could be used to circumvent certain safeguards. However, the company argued that the issue was already known, minor in scope, and comparable to capabilities available in other frontier AI systems.
“We believe this is a misunderstanding and are working to restore access as soon as possible,” Anthropic said in a statement.
Mythos 5 Raised International Attention
Anthropic first introduced Mythos 5 in April through a restricted release to a small number of organizations. The company described the model as possessing advanced cybersecurity capabilities, including the ability to identify and exploit software vulnerabilities.
The limited launch generated concern among governments and critical infrastructure operators, particularly in Europe, where organizations were not initially granted access to the technology.
Policymakers warned that highly capable cyber-focused AI systems could alter the cybersecurity landscape by enabling more sophisticated attacks while simultaneously strengthening defensive capabilities.
The restricted release also intensified discussions around technological sovereignty within Europe, with lawmakers urging greater investment in domestic AI and cybersecurity capabilities.
European Institutions Sought Access
The launch of Mythos 5 prompted questions from members of the European Parliament regarding the European Union’s preparedness for increasingly capable AI-driven cyber threats.
Lawmakers called on the European Commission to outline plans for developing sovereign AI cybersecurity capabilities and strengthening resilience against potential future risks.
Following discussions between Anthropic and European authorities, the EU cybersecurity agency, ENISA, was reportedly granted access to the model in early June.
Growing Focus on AI and Cybersecurity
The incident highlights increasing regulatory scrutiny of advanced AI models with cybersecurity applications.
Industry observers have noted that several major AI developers are pursuing similar technologies. Anthropic’s competitors, including OpenAI and other frontier AI companies, have also introduced models designed to support cybersecurity research, vulnerability detection, and software analysis.
Anthropic maintains that the capabilities cited by the US government are not unique to Fable 5 and remain broadly consistent with functionality already available in other leading AI systems.
The company warned that applying restrictive standards based on limited security bypasses could significantly affect the deployment of future frontier AI models across the industry.
For now, access to Mythos 5 and Fable 5 remains suspended as Anthropic works with US authorities to clarify the concerns and seek a resolution that would allow the models to return to service.
Source: DataMagz
Tidak ada komentar: