Adobe, Nvidia Join 100+ Experts in Urgent AI Security Letter

More than 100 cybersecurity experts and leaders from major tech companies including Adobe and Nvidia have signed a letter to the U.S. government, warning that Anthropic's Mythos AI models are alarmingly effective at identifying software vulnerabilities and weaponizing exploits. The letter, revealed on June 23, 2026, comes amid growing tensions between the AI company and the Trump administration over national security risks.

Mythos: A Double-Edged Sword

Anthropic's Mythos model, already known for its advanced reasoning capabilities, has demonstrated a troubling proficiency in cybersecurity. According to a U.S. official who spoke on condition of anonymity, Mythos identified vulnerabilities in highly sensitive government systems within hours during a test conducted under Project Glasswing—an Anthropic initiative that partners with intelligence agencies to assess AI risks. Senator Mark Warner (D-VA) confirmed the findings, stating that the tool breached nearly all classified systems in hours, not weeks.

The signatories, which include chief information security officers from Adobe and Nvidia, argue that while Mythos could be a powerful tool for defense, its offensive capabilities pose an unprecedented threat. The letter urges the government to establish strict protocols for the development and deployment of such models.

Regulatory Crossroads

The warning arrives at a critical juncture. The Trump administration recently issued a directive restricting access to a limited version of Mythos, citing cybersecurity fears. This followed an executive order mandating federal vetting of advanced AI systems before public release. However, Anthropic has raised concerns about military use of its technology, creating friction with the administration.

Meanwhile, the Commerce Department rescinded Biden-era chip export restrictions, authorizing the transfer of 35,000 Nvidia Blackwell chips to UAE-based G42—a move that some senators argue raises conflict-of-interest questions given the UAE's $500 million investment in Trump's crypto venture. The intersection of AI, national security, and geopolitics is becoming increasingly complex.

Industry Response and Future Implications

The letter represents a rare moment of consensus among tech giants and security experts. Adobe and Nvidia, both heavily invested in AI, are signaling that the risks of unregulated AI development cannot be ignored. The signatories call for a balanced approach that fosters innovation while safeguarding critical infrastructure.

Project Glasswing, which brought together companies like Microsoft and Google, aims to preemptively secure software from potential AI-driven attacks. The project's findings underscore the urgency: Mythos can not only find flaws but also automate exploit generation, a capability that could overwhelm current defense systems.

As the debate over AI regulation intensifies, the letter serves as a stark reminder that the technology's dual-use nature demands careful oversight. With the 2026 midterm elections approaching, lawmakers face pressure to act—but partisan divisions may stall meaningful legislation.

1. Over 100 cybersecurity experts from Adobe, Nvidia, and others warn about Anthropic's Mythos AI's exploit capabilities.
2. Mythos breached classified U.S. systems in hours during Project Glasswing tests.
3. The Trump administration has restricted access to Mythos amid growing tensions with Anthropic.
4. Industry leaders call for balanced regulation to mitigate risks without stifling innovation.

Sources: CNBC, CryptoNews