SAN FRANCISCO, United States — A powerful new artificial intelligence model developed by Anthropic has triggered a global security scramble, raising urgent concerns among governments and financial institutions about the future of cybersecurity and geopolitical stability.
The model, known as Mythos, has been described by its creators as capable of identifying and exploiting hidden vulnerabilities in critical systems, including banking networks, power grids, and government infrastructure.
Its release — tightly restricted due to safety concerns — has ignited debate over who should control access to such transformative and potentially dangerous technology.
Anthropic disclosed that it is working with a limited group of more than 40 organizations to assess and mitigate risks posed by the model, naming 11 U.S.-based partners, including major technology firms.
Britain remains the only country outside the United States granted direct access so far, underscoring widening global inequalities in AI capabilities.
The Bank of England has already sounded the alarm, warning that the technology could “crack the whole cyber-risk world open.”
European regulators have also moved quietly, with the European Central Bank querying financial institutions about their preparedness against advanced AI-driven cyberattacks.
Canada’s finance ministry likened the potential disruption to a strategic chokepoint crisis, highlighting the perceived scale of risk.
“This represents a step up in AI cyber capabilities,” Britain’s AI minister said following an independent evaluation by the country’s AI Security Institute, which confirmed that Mythos can execute complex cyberattacks beyond the reach of previous systems.
Anthropic chief executive Dario Amodei recently met with U.S. officials amid rising concern within government circles about the model’s implications for national security.
While the company has emphasized its cautious rollout, it acknowledged intense global demand for access, with governments and corporations seeking insight into the model’s capabilities.
“We are keeping access limited due to safety and security considerations,” the company said, adding that many organizations lack the technical capacity to responsibly handle such advanced tools.
The situation has exposed a critical gap in global governance. Unlike nuclear or chemical weapons, there is no international treaty, inspection regime, or standardized protocol governing the development and deployment of frontier AI systems.
“Access to foundational AI models is becoming geopolitical,” said Eduardo Levy Yeyati, an economist and advisor on AI policy. “Governments can no longer afford to treat this as a purely technological issue.”
In geopolitical terms, Mythos has amplified concerns among U.S. rivals. While official responses from China and Russia have been limited, analysts say both countries are closely monitoring developments.
Some observers in Moscow have gone as far as comparing the model’s potential impact to that of nuclear weapons, reflecting heightened anxiety over technological asymmetry.
Within China’s AI research community, the development is being viewed as a significant wake-up call.
Experts note that restrictions on advanced semiconductor exports have already slowed China’s progress in cutting-edge AI systems, potentially widening the gap with U.S. firms.
Meanwhile, European officials have expressed frustration over limited access.
Despite multiple engagements with Anthropic, the European Commission has yet to secure direct use of the model, citing unresolved concerns over data sharing and safeguards.
Germany’s cybersecurity agency has described the technology as marking a “paradigm change” in cyber threats, warning that traditional defense mechanisms may be insufficient against AI-driven attacks of this scale.
Anthropic has also disclosed that it is investigating reports of unauthorized access to a version of Mythos, raising further questions about containment and security controls.
The company estimates that similar AI capabilities could become widely available within 18 months, compressing the timeline for governments and industries to upgrade their defenses.
