Japan's 'Off-White Hackers': An AI Penetration Testing Team at IPA, and the 'Eve of a Cyber Emergency'

Key takeaways

• Japan's minister of economy, trade and industry, Ryosei Akazawa, said on October 23, 2026 that he has ordered a team of "off-white hackers" to be set up at IPA, the government's IT security agency, in November.
• Its job, in his words, is to attack systems harder than attackers do, using high-performance AI to run thorough penetration tests and find weaknesses first. He said the people will be chosen with attention to their trustworthiness.
• It is one answer to the breach wave. Two days earlier, a ruling Liberal Democratic Party panel called the situation the "eve of a cyber emergency" and asked for a cyber response headquarters led from the Prime Minister's office. Prime Minister Sanae Takaichi said on October 23 that the government will build that kind of structure.
• The announcement leaves the practical questions open: whose systems the team will test, under what legal basis, how big it will be, and whether its findings will be shared. It is also unclear why the team sits at IPA, under one ministry, when the government's cyber command center is the National Cybersecurity Office.
• In our view, attacker-minded testing is a sound idea, but "off-white" is the wrong word for it: what makes ethical hacking legitimate is permission, not the shade of the hacker. And this wave was driven mainly by known flaws and exposed settings. Finding them was never the hardest part. Getting thousands of organizations to know what they run and fix it is.

What the minister announced#

Minister Akazawa opened his regular press conference on October 23 with the plan. His reasoning, from the ministry's transcript:

• Attackers in recent incidents may be using AI to search widely for vulnerabilities. Defenders need to use AI too, and act first.
• Defense is structurally at a disadvantage: it is built around attacks it expects, an attacker needs only one hole while defenders must cover everything, and attacks tend to come from the unexpected.
• So the answer is to think like the attacker: use high-performance AI on the defensive side and thoroughly run penetration tests, simulated attacks that check a system's weak points.

To do that, he said, the government will gather "top gun" people who can attack beyond what attackers do and find vulnerabilities. He noted that the term "off-white hacker" seems to have come into use recently, and described them as people who keep attack in mind while making the defense succeed as a whole. The team will be set up within November at the Information-technology Promotion Agency (IPA), which the ministry oversees, "while taking care to ensure the reliability of the people". Later in the same press conference he called it a "white hacker team".

The minister also said:

• On October 22, working with the National Cybersecurity Office, the ministry warned companies about the PHP flaws through about 1,000 industry associations. The websites of three bodies under the ministry, including the export credit agency Nippon Export and Investment Insurance, were in maintenance mode that night.
• Information that companies have sent to the ministry about attacks, which it requested on October 9, is being passed to the National Cybersecurity Office. He declined to describe the attack methods, saying details could lead to further attacks.
• For small businesses, he pointed to the government-backed Otasuketai security support service, whose cost can be covered by a digitalization subsidy, and to the supply chain security rating scheme starting next March.

What "off-white" means#

"Off-white" is not an official category. In Japanese media, it has recently been used to describe a hacker whose work is protecting organizations but whose attack skills are on a par with real attackers, a step away from the clean image of the "white hat". The minister used it the same way: someone who attacks so that the defense succeeds. Later in the same press conference, he went back to calling it a "white hacker team". We think the word is a poor choice; see our analysis.

The bigger picture: "the eve of a cyber emergency"#

The announcement came in a week of political pressure.

The LDP's emergency proposal. On October 20, the LDP's National Cybersecurity Strategy Headquarters, chaired by former digital minister Masaaki Taira, compiled an emergency proposal, which the party approved on October 22. According to Jiji Press, FNN and Business+IT, it:

• calls the situation "a serious crisis that could be called the eve of a cyber emergency, shaking Japan's national security at its root"
• asks the government to set up a "cyber response headquarters" led from the Prime Minister's office as the command center, and to speed up measures across all ministries
• warns that AI could make "saturation attacks", many automated attacks at once, a reality, and asks for budget and laws to be prepared before then
• sets three directions: raising the security of companies that hold large amounts of important data, putting AI and next-generation cryptography to use, and building an active defense network in peacetime
• asks the government to consider making the My Number card compulsory as a safer form of identification, after this autumn's leaks of driver's license images, and to promote smartphones with biometric authentication
• calls for a direct line to major US AI developers, and for programs to hire private-sector experts into government with new pay rules

The Prime Minister's response. Responding to the proposal on October 23, Prime Minister Takaichi said that, as the proposal puts it, the country is on the "eve of a cyber emergency", and that the public's anxiety is "on a completely different level from before". She said the National Cybersecurity Office would be the command center, and that the government would build the necessary structure led from her office and speed up ministries' measures, according to ANN and NHK.

What we don't know yet#

The ministry has not said:

• How many people the team will have, how they will be recruited, or what it will cost
• Whose systems it will test: government systems, companies that ask for it, critical infrastructure, or something else
• What legal basis it will use if it tests systems without their owners' explicit consent
• Which AI it will use, and whether data from tests will go to outside AI services
• How its findings will be shared with the organizations affected and with everyone else
• Why it is being set up at IPA rather than under the National Cybersecurity Office, and how it will relate to the office and to the cyber response headquarters the LDP wants

Our analysis#

A word that sends the wrong signal#

In our view, "off-white" is the wrong word for a government team, and it matters more than it might seem.

In security, white, black and grey hats are not about skill. Any good penetration tester has attack skills on a par with real attackers; that is the job. What separates an ethical hacker from a criminal is permission and rules: a client who agrees to the test, a defined scope, and findings that go back to the owner. The legitimacy of the whole profession rests on that line. "Off-white" suggests that the line can be smudged a little for the sake of effectiveness.

That sends the wrong signal to three audiences. Companies have to trust a team before they let it attack their systems, and "not quite white" is not reassuring. Candidates may read it as a promise of fewer constraints, which is exactly the wrong reason to join. And it sits awkwardly with the minister's own point about the reliability of the people he wants to hire. Japan's security community has spent years establishing that ethical hacking is a legitimate profession; a government label that blurs it does not help. The minister himself switched back to "white hacker" later in the same press conference. We hope that is the name that sticks.

The idea behind the word, testing with an attacker's mindset and tools rather than with a checklist, is a good one. It does not need a new color.

Why IPA, not the NCO?#

The National Cybersecurity Office (NCO) was set up in the Cabinet Secretariat in July 2025, under the active cyber defense law, to lead Japan's cyber policy. On the same day as the minister's announcement, the Prime Minister named it the command center for the response to this wave, and the LDP wants a headquarters led from her office on top. The information companies send to METI about attacks is already being passed to the NCO.

Yet the new team is being set up by one ministry, METI, at IPA, an agency that METI oversees. There are practical reasons this might make sense: IPA is an operating agency with technical staff and the ability to hire and contract, while the NCO is mainly a coordinating office; and METI's remit is industry, which is where most of this autumn's victims are. But the minister did not explain the choice.

In our view, the risk is fragmentation. Japan already has cyber capabilities spread across agencies: NICT under the communications ministry, the police, the Self-Defense Forces, and IPA itself. If each ministry builds its own team, the questions that matter, such as which systems get tested first, who sees the findings and how they reach other sectors, end up decided ministry by ministry. Whatever its home, the team's priorities and its results should run through the command center the government has just named. How the IPA team, the NCO and the proposed headquarters fit together is something the government should spell out before November.

Whose systems?#

Attacking a system you do not own, even to help, is a crime in Japan under the Unauthorized Access Prohibition Act unless the owner agrees. Penetration testing works because the client gives permission. When the government wanted to try weak passwords on internet-connected cameras and routers in its NOTICE project, it first had to amend the law governing NICT, the national research institute that runs it, and the project still needs a plan approved by the minister.

In our view, that makes the scope of the team the most important open question. A team that tests only government and IPA-related systems, or companies that ask for it, is easy to set up but reaches few of the organizations hit this autumn. A team that tests anyone's internet-facing systems, the way attackers do, would need a legal basis that does not exist today. The minister's words, "attack beyond what attackers do", fit the second. The likely reality in November is the first.

Would it have stopped this wave?#

The attacks behind Japan's autumn breaches were not mainly about undiscovered flaws. JPCERT/CC's advisory describes scanning for known vulnerabilities, abusing internal APIs, a known Metabase flaw, web shells and AI pentest tools aimed at exposed configuration and backup files. Times Car was a web system vulnerability followed by credentials found inside; Arcadia was a path traversal and a secret in a config file.

Penetration testing would find many of those. But the gap was rarely that nobody could find them. It was that organizations did not know everything they had exposed, did not patch it, or kept secrets where an intruder could read them. A government team can show a company its holes. Closing them still needs the company's own asset inventory, patching and secrets management, and that is where we would want the policy effort to go too.

Where AI on the defensive side has already made a difference this month is in finding new flaws in widely used software: the PHP vulnerabilities were found through AI-assisted scanning before attackers used them. In our view, a national team would add most value doing that kind of work on software that many Japanese organizations share, such as local government systems, popular business packages and the platforms behind this autumn's breaches, rather than testing companies one at a time.

One team, thousands of targets#

Every breach we have covered this autumn was a different organization, and behind them are thousands more with similar exposure. No single team at IPA can test them all. In our view, the measure of success will be what the team produces for others: methods, AI tooling and findings that private testers, vendors and companies' own security teams can reuse, and help through the existing Otasuketai service for small businesses that cannot afford testing at all.

Sharing, not just finding#

The same press conference showed the tension. The minister said the government is collecting attack details from companies, but will not describe them because that could help attackers. Yet the most useful thing published this month was the opposite: JPCERT/CC's advisory with attack patterns and IP addresses, which defenders can act on. As we argued in our coverage of ARTEX, attackers already know their methods; defenders are the ones missing them. A team that finds weaknesses should be judged partly on how much of what it learns reaches the people who need to fix them.

Vetting the team#

The minister's remark about the reliability of the people is worth taking seriously. The skills the team wants are the same skills used in attacks, and the AI tools involved, such as ARTEX, are freely available to both sides. Japan has had a security clearance system for economic security information since 2025, but the minister did not say whether it will apply, or how candidates from the private sector will be vetted. Clear rules of engagement and vetting, published in advance, would do more for the team's credibility than any name.

What this means for readers#

• If you run security for a Japanese organization, do not wait for the government team. Commission your own penetration tests, including AI-assisted ones, and start with an inventory of what you expose to the internet.
• If you are a penetration tester in Japan, watch for how IPA recruits. The vetting rules will say a lot about what kind of team this will be.
• If you sell software used widely in Japan, expect more attention, from AI tools on both sides.
• Watch for the cyber response headquarters, any law change that would let the team test systems without consent, and whether the team publishes what it finds.

Japanese terms at a glance#

Japanese	Reading	Meaning
オフホワイトハッカー	ofu howaito hakkā	"Off-white hacker": a defender with full attack skills
ホワイトハッカー	howaito hakkā	White hacker, ethical hacker
ペネトレーションテスト	penetorēshon tesuto	Penetration test
サイバー有事前夜	saibā yūji zen'ya	"The eve of a cyber emergency"
サイバー対処本部	saibā taisho honbu	Cyber response headquarters (proposed)
飽和攻撃	hōwa kōgeki	Saturation attack
情報処理推進機構（IPA）	jōhō shori suishin kikō	Information-technology Promotion Agency
不正アクセス禁止法	fusei akusesu kinshi hō	Unauthorized Access Prohibition Act