
Google to Give Japan Early Access to Frontier AI Cyber Defense Models
Google will give Japan early access to its frontier AI cyber models, Nikkei reported Nov. 16, as Tokyo approves a ¥10 billion package after mass breaches.
- By
- Grace Kim
- Filed
- Channel
- AI & Compute
- Read
- 3 min read
Google will give Japan's government and major infrastructure providers early access to its latest AI model for cybersecurity, Nikkei Asia reported on Nov. 16. Google Cloud CEO Thomas Kurian was in Japan that day for talks with government officials, the report said.
Nikkei described the move as an expansion of an existing Google cyberdefense support program. The report did not name the model or specify which agencies and companies would take part. Google had published no announcement of its own in the sources reviewed, so the arrangement rests on a single outlet's account.
How does Google's early-access program work?
Google DeepMind operates the Fairwind program. It gives vetted "high-priority defenders" early access to Google's "Gemini for cyber" models, used alone or with CodeMender, the company's code security agent that finds and patches vulnerabilities.
Access follows a strict priority order:
- Governments and national cyber authorities come first.
- Critical infrastructure operators in healthcare, telecom, energy and finance come next.
- Core technology platforms follow.
Google says it works with "thousands of partners globally." The Fairwind program page does not mention Japan, which suggests the new arrangement extends an existing mechanism rather than creating one.
Why is Japan seeking privileged access?
Japan faces a sharp rise in cyberattacks. Prime Minister Sanae Takaichi told officials in late October to devise steps to strengthen cyber defenses. The government then approved a ¥10 billion emergency cyber package after dozens of companies and agencies were breached. Last week, the City of Osaka warned that more than 1 million email addresses had leaked through a vendor.
Japan has also pursued wider access to advanced AI models from Anthropic and OpenAI. On Nov. 5, Nikkei reported that Tokyo would ask US AI developers for earlier access to information on undisclosed software vulnerabilities, so agencies and infrastructure operators could respond faster, according to Reuters.
Google already works with Japan's National Institute of Information and Communications Technology and the Digital Agency through its Japan Cybersecurity Initiative, according to Google's Japan blog.
What is the wider pattern among AI labs?
Frontier labs have been gating their strongest cyber-capable models through controlled early-access programs rather than public release. Google's offer to Japan fits that pattern, giving a state and its operators privileged tools before general availability.
The timing is notable. Earlier on Monday, the same day Kurian met Japanese officials, Google said it had caught hackers running an AI agent that found and exploited software flaws with no human involved.
What remains unconfirmed?
Several details are unresolved. The model involved is unnamed, the participating agencies and companies are unspecified, and Google has not issued its own confirmation. If the arrangement follows the Fairwind template, Japan's government and cyber authorities would gain access first, with infrastructure operators behind them.
The deal signals that early access to frontier cyber AI models is becoming an instrument of alliance and national defense policy, and more governments are likely to negotiate similar terms as attack volumes rise.
Source: mlq.ai
More from Grace Kim
Show full bio
Market editor covering people, launches and day-to-day practice in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.
1,035 articles
Related articles
google-launches-gemini-4-argon-limits-access-to-trusted-cyber-defenders-8e96a2ca
Google Launches Gemini 4 Argon, Limits Access to 'Trusted Cyber Defenders'
google-tests-its-first-ai-satellite-in-orbit-f45590f8
Google Tests Its First AI Satellite in Orbit
google-org-puts-2-3-mn-into-froncort-ai-for-disease-surveillance
Google.org Puts $2.3 Mn Into Froncort.AI for Disease Surveillance
google-adds-15b-for-malaysian-data-centers-and-grid-power
Google Adds $15B for Malaysian Data Centers and Grid Power



