June 19, 2026, (Inside AI) — Anthropic has kept its most advanced Mythos AI model accessible to a select group of early testers, even after a U.S. government order forced the public version offline. The preview, limited to about 200 organizations under the Glasswing program, remains active for some firms.
A Split Decision on Access
Bloomberg News reported Thursday that companies like Dragos and Cisco Systems confirmed they still have access to Mythos Preview. The model had flagged thousands of software vulnerabilities before the clampdown.
The public, less powerful version of Mythos was disabled after Washington issued export control directives. The order suspended access for all foreign nationals, citing national security risks.
Anthropic has not explained how it decides which Glasswing members keep access. The company, Cisco, and Dragos did not respond to Reuters’ requests for comment outside business hours.
Europe’s Cybersecurity Watchdog Left Out
The European Union Agency for Cybersecurity (Enisa) had been invited to join Glasswing before the U.S. block. On Friday, it was told it would no longer get access.
Enisa was not immediately available for comment. The exclusion highlights the geopolitical tensions shaping AI governance.
Anthropic’s Fractured Government Ties
Anthropic’s relationship with the U.S. government soured earlier this year. The IPO-bound company refused to let the military use its AI for domestic surveillance and fully autonomous weapons.
That stance set the stage for the current restrictions. The export control order appears to be a direct response to those ethical boundaries.
What the Selective Access Means
The Mythos Preview’s continued operation for some suggests a tiered trust system. It raises questions about transparency and fairness in AI deployment.
Security researchers note that the model’s vulnerability-finding capability is a double-edged sword. It could strengthen defenses or enable exploitation if mishandled.
Anthropic’s silence on the criteria fuels speculation. Industry observers wonder if access depends on contractual agreements or strategic partnerships.
Looking Ahead
The situation underscores the growing friction between AI innovation and national security. As models become more powerful, governments are likely to tighten controls.
Anthropic’s next moves will be closely watched, especially with its IPO on the horizon. The company must balance ethical commitments with regulatory pressures.
Shamil has a low tolerance for AI that only works in controlled demos. He covers agentic systems and generative AI with the skepticism of someone who has seen too many “revolutionary” tools quietly disappear six months after launch.
