JAKARTA - Amid the official government's stance of including Anthropic on the supply chain risk list, the latest report says the National Security Agency (NSA) still uses the company's cutting-edge AI model, Mythos Preview.
This situation presents a policy paradox: on the one hand, formal restrictions are imposed, but on the other hand, operational needs actually encourage the use of the same technology. A number of sources say the NSA is among the approximately 40 organizations that have exclusive access to the model.
The move comes after the Donald Trump administration in February 2026 ordered government agencies to stop using Anthropic services. The policy was triggered by a deadlock in negotiations, after Anthropic refused to loosen restrictions on using their AI for autonomous weapons or domestic mass surveillance.
However, behind this policy, the need for high-level cyber capabilities seems difficult to ignore. Mythos is known to have advantages in digital security analysis, including the ability to identify system vulnerabilities and simulate potential exploits - features that are very valuable for intelligence agencies.
This model was even released in a limited way only to a verified group, given the potential "offensive cyber capability" it has. In practice, most early users use Mythos to scan their own systems, looking for vulnerabilities before they are exploited by others.
The signal of the US government's interest in this technology was also seen from a meeting at the White House. Anthropic CEO, Dario Amodei, reportedly met with White House Chief of Staff Susie Wiles and Secretary of the Treasury Scott Bessent. The meeting discussed the possibility of using Anthropic technology outside the military sphere.
Although President Trump admitted not knowing the details of the meeting, officials described the discussion as constructive. The focus is on finding a point of convergence between the need for technological innovation and national security constraints.
This phenomenon confirms a new reality: in the era of generative AI, the line between risk and need is getting thinner. Governments are faced with a dilemma - holding back technology for security, or using it to maintain strategic advantage.
It's not just the United States that is moving fast. In the UK, the UK AI Security Institute has also confirmed that it has accessed Mythos, signaling a global race to dominate high-powered AI technology.
If this report is accurate, then the Mythos story is not just about one AI model - but the beginning of a new round of cyber competition, where even official bans can waver in the face of strategic needs.
Follow VOI Whatsapp Channel
The English, Chinese, Japanese, Arabic, and French versions are automatically generated by the AI. So there may still be inaccuracies in translating, please always see Indonesian as our main language. (system supported by DigitalSiber.id)
