Five Eyes agencies urge faster cyber defenses as AI hacking risks grow
The intelligence-sharing bloc said advanced AI could sharpen offensive hacking within months and called for faster patching and defensive use of AI.
By Sofia Marchetti · World Affairs Correspondent
3 min read
The Five Eyes intelligence partners have warned that advanced artificial intelligence systems could make offensive hacking more powerful within months, adding urgency to efforts to secure government and private networks. In a three-page statement issued Monday, officials from the United States, United Kingdom, Canada, Australia and New Zealand said frontier AI could reshape both cyberattacks and cyber defense.
The alliance said the timetable for the threat was measured in months rather than years. Reuters reported that the statement offered few technical details and largely reinforced standard cybersecurity practices, including rapid fixes for flawed software and limits on exposing systems to the internet when access is not needed.
The warning shows growing official concern about specialized AI models that could help users carry out complex cyber operations more quickly, Reuters reported. The models cited in that report include Anthropic’s Mythos and OpenAI’s GPT-5.5-Cyber, which officials and researchers have said may expand the ability of users to conduct damaging intrusions.
Agencies push basic defenses and AI tools
The Five Eyes statement urged defenders to use AI on their own networks, including to find weaknesses earlier and respond faster when incidents occur. The agencies said advanced models could affect both sides of cybersecurity, giving attackers new tools while also offering defenders ways to spot and contain threats.
The guidance also pointed agencies and organizations back to core security work. According to Reuters, the recommendations included patching defective software quickly and avoiding unnecessary online exposure for systems, steps long treated as basic cyber hygiene.
The statement was co-signed by the US Cybersecurity and Infrastructure Security Agency, known as CISA, Reuters reported. CISA had already tightened its own timetable for federal agencies to address serious vulnerabilities, cutting the deadline to three days and citing threats linked to AI.
Model access draws government scrutiny
The Five Eyes warning followed action involving Anthropic’s Mythos model earlier this month, according to Reuters. Anthropic disabled a version of Mythos after the US government ordered the company to suspend access for foreign nationals over alleged national security concerns.
Reuters reported that concern has increased around systems designed or adapted for cyber tasks because they may let users complete difficult hacking operations at speed. The Five Eyes agencies did not lay out a detailed public account of the capabilities they fear, but their statement framed the issue as urgent.
The Five Eyes alliance is an intelligence-sharing partnership among the US, UK, Canada, Australia and New Zealand. Its latest cyber warning puts advanced AI alongside longstanding security failures such as unpatched software and unnecessary exposure of systems, signaling that officials want faster defenses before those tools become more widely available.
This story draws on original reporting from Al Jazeera.