World-wide AI tools in cyber security: preparing for possible disruptions

Views: 24

In April 2026, the US’ Anthropic disclosed a new frontier AI model (Claude Mythos), able to find and exploit security vulnerabilities in software better than “any most skilled humans”. The company noted, for example, that the model had uncovered some flaws that had gone undetected for several decades in a secure operating system used to run firewalls that guard sensitive national networks. Mythos, it seems, is rapidly changing traditional approaches to cybersecurity and national/global security, the issues vital for the European digital agenda as well. 

Background
Presently, two US digital companies – Anthropic and OpenAI – have publicly demonstrated AI models advanced enough not only to detect flaws in software far beyond the human capacity, but able to manage the use of these models to defensive cybersecurity operations.
However, some experts note that Anthropic’s Mythos facilities just “constitute marketing” approach, showing the leading commercial software companies around the world the most progressive AI model. Thus, it claimed that the model is outpacing elite digital gurus in faster digital solutions to complicated security challenges.
Still a general question is: how the global major powers can “defend from the AI-powered cyberwarfare”? That is, e.g. the US may only have about 9-12 months to protect its critical infrastructure from AI-powered attacks, notes M. Sulmeyer, Professor at Georgetown University, previously serving as the US assistant secretary of defense for cyber policy.
For example, China is already acquiring these AI capabilities and may be closer to developing its own Mythos-type facilities making advanced AIs sharpen China’s “credible threat to deter a fight over Taiwan before it begins by disrupting the island’s critical infrastructure and that of any nation that seeks to defend it”.
Source and citation from: M. Sulmeyer. “When China Gets Its Own Mythos. Preparing for an AI Cyber Crisis”. – Foreign Affairs. July 2026. https://www.foreignaffairs.com/china/when-china-gets-its-own-mythos.

Generally, attacks on critical infrastructures (like energy, water purification plants or pipelines) were rare because they were time- and resource-intensive. Mythos has already demonstrated that it can find and exploit flaws with minimal human direction; thus, such a model could generate many damaging compromises automatically and quicker. The new models could allow for a kind of cyberattack different from any prior ones in scale and complexity. The general opinion is that most advanced AI models could soon make widespread physical compromise something a single adversary can execute.

Global digital competition: US vs. China
As AI models advance, they become powerful tools for building even more capable AI. It should not come as a surprise that American AI companies are using their own models to build the next generation of models, but it is easy to miss the fact that the United States’ chief competitors are also using top U.S. models to engineer the next version of their own models, as well.
Understanding the danger, Anthropic has not released Mythos openly, instead putting it into the hands of defenders first by restricting access to a coalition of technology, finance and open-source organizations, an effort it calls Project Glasswing.
This was a notable decision: a company chose to withhold its most powerful product from broad release because the product could be turned, with little modification, into a weapon. Eventually, Anthropic offered the Fable model to the public: a product using the same underlying model but with enhanced safety guardrails.
It is expected that within weeks the Anthropic would launch Project Glasswing, an OpenAI made GPT-5.5-Cyber available to vetted security teams through a program it called Trusted Access for Cyber. The company revealed that the early results were stunning: Microsoft, using an internal scanning system built by applying multiple AI models, disclosed 16 previously undetected flaws in the Windows networking stack. And the cybersecurity firm Palo Alto Networks, testing the new models against its own products, reported finding far more vulnerabilities in a single scan than it typically discloses in a month.

= China is not only building domestic alternatives to American AI models. Chinese actors are making efforts to gain illicit access to the most powerful export-controlled AI semiconductor chips. Over the past year, the US Department of Justice has prosecuted multiple large chip-smuggling networks; however, smuggling alone cannot deliver the compute needed to train advanced models. But just as China is building its own models, it is building its own chips.

= The United States are urgently learning the lessons of the long-running Chinese campaign to steal American intellectual property. It took too long to stand up protections such as the Defense Department’s threat sharing with cleared contractors. In 2014, the US federal indictment accused Chinese military officers of hacking American firms, including the nuclear-reactor manufacturer Westinghouse, as early as 2006, but that protective program was not formalized until 2012. Presently, the digital companies throughout the AI stack represent a high-value foreign intelligence targets; hence, the United States should not repeat that delay and, instead, swiftly strengthen counterintelligence support and other efforts to check the Chinese playbook, acknowledged the US Administration.

In June 2026, the US President signed an executive order creating a voluntary framework giving the government up to 30 days of early access to new frontier models before their broader release and directing federal IT networks to make some security enhancements.
Experts say that this was “a useful first step’: but the government must mount a much fuller offense. The measures that follow aim to preempt attacks and prevent harm; but some intrusions will get through, so the US governance must also build the resilience to absorb, contain and recover from the compromises it cannot stop.

Bottom-line
A wide range of world-wide digital technologies is not only reshaping traditional international relations; it systematically institutionalizes digital geopolitics as an emerging sphere of global competition. Hence, the “combination” of digital technologies (including the AIs and ICTs, big data and infrastructures, etc.) consequently multiplies the digital geopolitics’ influence on global governance and power competition.
Presently, the “leading status” of digital technology “landscape” between the US and China greatly impacts both the two sides’ technology’s competition and the global digital power architecture. The resolution of this digital competition could not only determine the basics of the future international order, but prevent the emergence of the “global digital cold war” and global digital security.
Basically, the digital geopolitics defines the US-China rivalry as a high-stakes race for technological primacy. Both superpowers compete intensely for dominance in such digital aspects as the AI models, semiconductor manufacturing, digital infrastructures and global data governance, fracturing the open internet into competing and highly sovereign spheres.
Reference to: https://www.researchgate.net/publication/394564259_Digital_Geopolitics_and_US-China_Competition

Among core aspects of the mentioned competition are:
= Semiconductors and hardware: advanced microchips are the choke points of the digital age; thus, the US uses export controls and multilateral restrictions (involving allies like the Netherlands and Japan) to block China’s access to extreme ultraviolet (EUV) lithography and advanced design tools. China responds with a heavy push for self-sufficiency and supply chain leverage in critical minerals.
= Artificial Intelligence: the AI models are viewed not merely as commercial tools but as a form of foundational infrastructural and military power. The US, e.g. relies on massive private-public domestic investments and deregulation to drive innovation, while China heavily backs state-directed tech ecosystems.
= Governance and data sovereignty: the US governing model promotes cross-border data flows and market-driven architecture, whereas China prioritizes strict data localization and national sovereignty over information. China extends its digital footprint globally through telecommunications, smart city infrastructure and fiber optics, building long-term technological dependencies in the so-called Global South.