Japan’s decision to revise its artificial intelligence (AI) strategy just six months after its initial rollout reveals the relentless pace at which global cyber threats are evolving.
The move comes as regulators grapple with the arrival of frontier AI models such as Anthropic’s Claude Mythos, which Anthropic says can find and exploit thousands of zero-day vulnerabilities once directed to do so, with little further human guidance needed.
Released on 19 June, the draft revision to Japan’s AI Basic Plan – first adopted in December 2025 – calls for stronger defences against cyber attacks that weaponise high-performance AI, an expanded role for the Japan AI Safety Institute (AISI) in evaluating advanced models, continued alerts to critical infrastructure operators, and inspections of vulnerabilities across key government systems.
The updated policy also prioritises support for technologies designed to detect AI-generated disinformation, as well as enhanced cooperation with ASEAN and other Global South countries. Tokyo is seeking public comment on the draft before targeting cabinet approval in July 2026.
“The draft revision’s emphasis on cooperation with foreign governments and AI developers, alongside its attention to cyber attacks that exploit advanced models, reflects the need to address an urgent problem that extends well beyond Japan,” said Takanori Nishiyama, senior vice-president of Asia-Pacific and country manager for Japan at Keeper Security.
Security analysts have warned that threat actors are already using AI to spearhead attacks such as phishing, malware development, password cracking and social engineering, while also exploiting weak credential hygiene and inadequate access controls – still the primary entry points for most cyber security breaches.
Nishiyama noted that advanced models capable of scanning systems, networks and code for vulnerabilities will only compound the danger, operating with a speed and scale no human analyst can match. “These models will accelerate the path from reconnaissance to exploitation faster than most security teams can detect, let alone respond to,” he added.
While Tokyo works to update its national defences, the immediate burden of resisting these high-velocity AI threats falls on enterprise security teams. Organisations must now operate at a faster pace, assuming public vulnerabilities will be weaponised within hours rather than weeks.
That requires organisations to implement automated update paths for internet-facing systems and treat security patches as urgent priorities rather than backlog tasks. Maintaining comprehensive logging and multi-factor authentication is also key to limiting lateral movement in the event of a breach.
To neutralise AI-driven attacks, Nishiyama called for organisations to move away from traditional perimeter defences towards internal containment, centred on privileged access management (PAM).
“PAM acts as an internal circuit breaker,” he explained. “By replacing vulnerable, always-on administrative accounts with just-in-time access and automated credential vaulting, PAM ensures an automated attacker finds no persistent rights or tokens to harvest.”
By isolating privileged sessions and applying user behaviour analytics to instantly terminate high-velocity anomalies, these systems can effectively trap exploits at their entry point before they spread across a company’s network.
Notwithstanding the arrival of these advanced technical threats, the fundamental tenets of cyber hygiene remain the most effective shield.
“The principles underpinning good cyber security have not changed significantly with the arrival of more capable AI,” Nishiyama said. “What has changed is the vital importance of applying them consistently.”