Blab AI
Five Eyes Warns of Rapid AI-Enabled Cyber Threats, Urges Immediate Action
The Five Eyes intelligence alliance – comprising the United States, United Kingdom, Canada, Australia and New Zealand – issued a joint statement on Monday warning that frontier artificial‑intelligence (AI) models could soon transform offensive hacking capabilities. The statement, released in Washington, said the timeline is not years but months.
The Five Eyes statement reiterated basic cyber‑security advice, such as patching software quickly and limiting the exposure of systems that are not essential. It also urged defenders to use AI to strengthen their own defenses, for example by identifying weaknesses sooner or responding more quickly to incidents.
The warning follows growing concerns about models that are specifically tuned for cyber‑operations. Anthropic’s “Mythos” and OpenAI’s “GPT‑5.5‑Cyber” are cited as examples of tools that can help users execute complex, potentially devastating hacks. According to Reuters, Anthropic was forced to disable a version of Mythos after the U.S. government ordered the company to suspend access to the model for foreign nationals over alleged national‑security concerns. The same month, the U.S. cyber‑defence agency CISA reduced the deadline for government officials to address serious digital vulnerabilities in their networks to three days, citing the threat posed by AI.
Anthropic’s Mythos model has been the subject of several high‑profile incidents. In mid‑June, the company suspended its Fable 5 and Mythos 5 models after the U.S. government raised security concerns. The suspension followed an authorised red‑team evaluation that reportedly showed the model could breach the National Security Agency’s classified systems within hours.
OpenAI’s GPT‑5.5‑Cyber, meanwhile, is marketed as a defensive tool that can identify flaws before they are exploited. The company has restricted access to the model to “critical cyber defenders” and has opened limited preview access to vetted European defenders under the EU Cyber Action Plan.
CISA’s new guidance reflects a broader shift in U.S. policy toward AI‑related cyber risk. The agency’s directive shortens the window for fixing vulnerabilities and stresses the need for rapid response. CISA also has issued guidelines for the responsible use of AI in critical infrastructure, emphasizing the importance of mapping, measuring and managing AI deployments.
The Five Eyes statement signals that intelligence agencies see frontier AI as a game‑changer for both attackers and defenders. While the statement does not detail specific capabilities, it underscores the urgency of preparing for a future in which AI can automate the discovery and exploitation of software weaknesses.
Defenders are advised to review their patch‑management processes, adopt AI‑assisted vulnerability scanning, and ensure that any AI tools used for defensive purposes are properly vetted and monitored. The statement also suggests that agencies should consider the potential for AI to accelerate the development of new attack vectors.
In the coming weeks, the Five Eyes partners are expected to coordinate further guidance on the use of AI in cyber‑operations. The U.S. government may also expand its restrictions on models that pose national‑security risks, while European regulators are likely to scrutinise the deployment of GPT‑5.5‑Cyber under the EU’s AI Act.
The current situation remains fluid. Anthropic and OpenAI are continuing to refine their models, and CISA is monitoring the impact of its new vulnerability‑remediation timelines. The intelligence community’s warning highlights the need for a coordinated, rapid response to an evolving threat landscape in which AI can shorten the window between vulnerability discovery and exploitation.
The Five Eyes statement reiterated basic cyber‑security advice, such as patching software quickly and limiting the exposure of systems that are not essential. It also urged defenders to use AI to strengthen their own defenses, for example by identifying weaknesses sooner or responding more quickly to incidents.
The warning follows growing concerns about models that are specifically tuned for cyber‑operations. Anthropic’s “Mythos” and OpenAI’s “GPT‑5.5‑Cyber” are cited as examples of tools that can help users execute complex, potentially devastating hacks. According to Reuters, Anthropic was forced to disable a version of Mythos after the U.S. government ordered the company to suspend access to the model for foreign nationals over alleged national‑security concerns. The same month, the U.S. cyber‑defence agency CISA reduced the deadline for government officials to address serious digital vulnerabilities in their networks to three days, citing the threat posed by AI.
Anthropic’s Mythos model has been the subject of several high‑profile incidents. In mid‑June, the company suspended its Fable 5 and Mythos 5 models after the U.S. government raised security concerns. The suspension followed an authorised red‑team evaluation that reportedly showed the model could breach the National Security Agency’s classified systems within hours.
OpenAI’s GPT‑5.5‑Cyber, meanwhile, is marketed as a defensive tool that can identify flaws before they are exploited. The company has restricted access to the model to “critical cyber defenders” and has opened limited preview access to vetted European defenders under the EU Cyber Action Plan.
CISA’s new guidance reflects a broader shift in U.S. policy toward AI‑related cyber risk. The agency’s directive shortens the window for fixing vulnerabilities and stresses the need for rapid response. CISA also has issued guidelines for the responsible use of AI in critical infrastructure, emphasizing the importance of mapping, measuring and managing AI deployments.
The Five Eyes statement signals that intelligence agencies see frontier AI as a game‑changer for both attackers and defenders. While the statement does not detail specific capabilities, it underscores the urgency of preparing for a future in which AI can automate the discovery and exploitation of software weaknesses.
Defenders are advised to review their patch‑management processes, adopt AI‑assisted vulnerability scanning, and ensure that any AI tools used for defensive purposes are properly vetted and monitored. The statement also suggests that agencies should consider the potential for AI to accelerate the development of new attack vectors.
In the coming weeks, the Five Eyes partners are expected to coordinate further guidance on the use of AI in cyber‑operations. The U.S. government may also expand its restrictions on models that pose national‑security risks, while European regulators are likely to scrutinise the deployment of GPT‑5.5‑Cyber under the EU’s AI Act.
The current situation remains fluid. Anthropic and OpenAI are continuing to refine their models, and CISA is monitoring the impact of its new vulnerability‑remediation timelines. The intelligence community’s warning highlights the need for a coordinated, rapid response to an evolving threat landscape in which AI can shorten the window between vulnerability discovery and exploitation.
