Japan

As for lawyers, in September 2025, the Japan Federation of Bar Associations (‘JFBA’) 一the national self-governing body for the legal profession in Japan一 set up an AI Strategy Working Group in June 2023 to examine AI’s implications for legal practice, its interaction with Article 72 of the Attorneys Act (which reserves legal practice to qualified lawyers), effects on the judiciary, and human rights impacts, and to consider drafting member-facing guidance. In September 2025, the working group published internal member-only guidance titled Points to Note Regarding the Use of Generative AI in Legal Practice. The working group guidance organises its concerns around four main themes: (1) the duty of confidentiality, (2) the risk of misinformation (hallucination), (3) the allocation of responsibility for AI-assisted work product, and (4) professional ethics more broadly. The guidance frames these obligations as existing professional duties that lawyers have always owed, now being reapplied to a new tool (rather than new rules applying to AI). This guidance does not represent the official position of the JFBA. The publication of the guidance was accompanied by explanatory seminars and other initiatives for members, aimed at helping practitioners digest the document.

Separately, the lawyers’ association Hoyukai has stressed the need for lawyers to be alert to the risk that generative AI may output incorrect statements of applicable law or cite non-existent caselaw, and cautions strongly against uncritical reliance on AI-generated answers to legal questions, requiring that any AI output be checked against sources and, where necessary, primary materials.

Criminal procedure rules

Code of Criminal Procedure (1942)

Although no rules or Supreme Court findings are expressly applicable to AI, several existing evidentiary doctrines under the Code of Criminal Procedure may affect the admissibility of AI-derived evidence introduced in criminal proceedings.

• Relevance: Relevance requires that evidence bear a logical connection to the fact to be proven, and that admitting it does not risk leading the court to misjudge its probative value. For scientific evidence, the Supreme Court requires that the underlying principle has a scientific basis, that the method is reliable, and that it is properly applied. Where an AI’s analytical process is a ‘black box’, its reliability may not be verifiable, risking that the evidence lacks relevance.
• Hearsay rule (Articles 320 through 328): Out-of-court statements are generally inadmissible unless a statutory exception applies. Even if a statement record is prepared or summarized using AI, it remains subject to the hearsay rule and must independently satisfy an exception.
• Illegally obtained evidence: Where the investigation involved a serious illegality, such as a breach of the warrant requirement (Article 218; Article 35 of the Constitution), evidence obtained is inadmissible. If AI-based investigative methods were used without, or beyond the scope of, a warrant, there is a risk that both the underlying data and any AI-derived analysis may be excluded.

The Criminal Digitalisation Act, which was passed in May 2025, amends the Code of Criminal Procedure and related statutes to:

• Enable the online submission of arrest and search warrant requests, with warrants issued and presented in electronic form (via tablet devices), removing the need for officers to travel to court in person, a journey that in some regional areas could take several hours each way;
• Digitise evidence documents (witness statements, investigative reports), enabling online disclosure and inspection by the defence, eliminating copying costs that can reach millions of yen;
• Expand video-link (remote) testimony to witnesses unable to attend due to illness and to bereaved family members participating under the victim participation system;
• Create new offences of falsification of public and private electronic records; and
• Introduce an electronic data provision order enabling investigators to compel telecommunications providers to transmit electronic data directly from their servers, with a confidentiality obligation (gag order) on the provider, limited after parliamentary amendment to one year.

The Criminal Digitalisation Act is expected to be fully implemented by the end of March 2027. Though the reform digitises procedure, it does not extend to the use of AI in decision-making within criminal proceedings.

The Criminal Digitalisation Act was passed despite concerns from opposition parties about the mass collection of personal information unrelated to the case under investigation. In response to these concerns, the final text includes a supplementary provision requiring investigators to ‘endeavour to avoid, to the extent possible, collecting personal information unrelated to the case.’

Japanese law contains no rule specifically addressing the exclusion of deepfake evidence. Evidence suspected of having been generated or altered by deepfake technology may nonetheless be excluded under the general evidentiary principles, either where the evidence lacks relevance, or, more fundamentally, where the proponent fails to establish the ‘authenticity’ that is a precondition for admitting the evidence.

Cybersecurity laws

Articles 23 and 66 of the APPI broadly require business operators and the heads of administrative organs to take sufficient and appropriate security management measures. These measures comprise:

• organizational measures (e.g., appointing a responsible officer, establishing handling rules, implementing an audit system);
• personnel measures (e.g., employee training and supervision, confidentiality obligations);
• physical measures (e.g., controlling access to handling areas, preventing theft of devices and media, secure data deletion on disposal); and
• technical measures (e.g., access control and authentication, anti-intrusion protections, encryption of communications).

The level of measures required depends on the nature and volume of the personal information involved and the risk posed by a potential leak.

The Cyber Response Capability Enhancement Act was enacted in May 2025 with the aim of enabling active cyber defence in Japan. The Act newly authorizes: (i) the acquisition and analysis of communications information to detect signs of cyberattacks; and (ii) access to, and neutralization of, devices involved in an attack, to be carried out by law enforcement officers and the Self-Defence Forces, with such neutralization measures expected to become available from 1 October 2026. The government is reportedly considering the use of advanced AI in carrying out these access and neutralization measures while maintaining human involvement in final decision-making.

Human rights

The use of AI tools in criminal proceedings in Japan must comply with the jurisdiction’s human rights framework, even though that framework does not explicitly refer to AI. Relevant human rights guarantees can be found in the Constitution of Japan 1946, which provides:

• Article 13: right to privacy/personality: ‘All of the people shall be respected as individuals. Their right to life, liberty, and the pursuit of happiness shall, to the extent that it does not interfere with the public welfare, be the supreme consideration in legislation and in other governmental affairs.’ The Supreme Court of Japan has recognised a right not to be photographed without justification known as the Kyoto-fu Gakuren case (Supreme Court, Grand Bench, 24 December 1969, Keishu Vol. 23, No. 12, p. 1625). This principle may apply to surveillance and profiling in which AI is used in a supportive capacity.
• Article 14: right to equality before the law: ‘All of the people are equal under the law and there shall be no discrimination in political, economic or social relations because of race, creed, sex, social status or family origin’. This guarantee may be engaged where AI tools used in criminal proceedings (such as risk-assessment or evidence analysis systems) embed or replicate biases present in their training data, resulting in disparate treatment of defendants or suspects.
• Article 32: right of access to court: ‘No person shall be denied the right of access to the courts.’ This right may apply to ensure that any future deployment of AI in judicial proceedings preserves the right to a hearing before a human judge.
  
• Article 35: requirements for lawful warrants: ‘The right of all persons to be secure in their homes, papers and effects against entries, searches and seizures shall not be impaired except upon warrant issued for adequate cause and particularly describing the place to be searched and things to be seized, or except as provided by Article 33.’ AI tools that generate evidence through automated, ongoing data collection would raise questions about whether a conventional warrant can satisfy the specificity requirement.
• Article 76: judicial independence and impartiality: ‘... All judges shall be independent in the exercise of their conscience and shall be bound only by this Constitution and the laws.’  Delegation of judicial power to AI would be constitutionally problematic. AI-assisted support (information organisation, reference to sentencing precedents) may be compatible with the current constitutional framework.

• Article 82: right to a public trial: ‘Trials shall be conducted and judgment declared publicly.’ Trial proceedings must be open. If AI’s internal processing logic is opaque, then explaining the basis of decisions to parties and the public becomes a challenge that could implicate the open-trial guarantee.

In addition to the constitutional rights mentioned above, human rights protections under the following international treaties to which Japan is a party may also be relevant to the proceedings, including the use of AI.

Japan is a party to the International Covenant on Civil and Political Rights (‘ICCPR’), ratified in 1979. Japan has not enacted implementing legislation to give domestic effect to the ICCPR as such, though it is generally understood that ICCPR provisions can, where sufficiently clear and unconditional, be directly applied (‘self-executing’) by Japanese courts. In a judgment of 3 February 1993, the Tokyo High Court held that Article 14(3)(f) of the ICCPR was self-executing and directly applicable, and on that basis quashed a lower court order requiring a foreign defendant to bear interpretation costs following conviction.

Japan also ratified the Convention on the Rights of the Child (‘CRC’) in 1994, and various domestic child-related laws were subsequently amended or enacted with reference to the convention principles. Article 12.2 guarantees a child’s right to be heard in judicial proceedings affecting them, and Article 40 sets out due process safeguards specific to children in conflict with the law (including the presumption of innocence and treatment appropriate to their age). These provisions may be engaged where AI-based risk-assessment or decision-support tools are used in proceedings involving juvenile suspects or defendants.

Moreover, Japan ratified the Convention on the Rights of Persons with Disabilities (‘CRPD’) in 2014 having already amended or enacted several domestic laws in preparation for ratification. Article 13 requires States to ensure effective access to justice for persons with disabilities, including through the provision of procedural accommodations at the investigative stage.

However, even though Japan is a party to these international treaties, it is generally considered that Japanese courts rarely invoke or rely on international instruments where a domestic provision already addresses the same issue. Accordingly, human rights protection against the use of AI in criminal and other relevant proceedings is likely to continue to rely primarily on constitutional protections and relevant domestic law, rather than directly on these international instruments.

Outlook

The role of AI in Japan’s justice system will likely expand gradually and cautiously, guided by centralised strategic planning at government level. Japan’s AI Strategic Headquarters holds regular strategy sessions, having held its fifth meeting in July 2026. The second Artificial Intelligence Basic Plan sets the government’s overarching policy framework for all AI research, development and utilisation across sectors.

In January 2026, the Regulatory Reform Promotion Council held a session addressing further clarification of AI use under the Attorneys Act, at which a Council member argued that legal technology and AI have become indispensable given deepening staffing shortages in the legal field, and pressed the Ministry of Justice to accelerate institutional clarification to keep pace with technological change. The Ministry of Justice reaffirmed its position that, in general, AI use by a lawyer in their own practice, or provision of legal AI services by a lawyer-operator provider, would not itself violate Article 72 of the Attorneys Act. The Council member pressed for a concrete roadmap ahead of the Council’s final regulatory reform report, suggesting further government clarification on the boundary between permissible AI-assisted legal practice and unauthorised practice of law.

CASES

As at May 2026, no Japanese criminal cases have been identified that directly address the admissibility, transparency, or legality of AI tools in criminal proceedings.