Japan AI appropriateness guideline (AI Promotion Act Art. 13)

Standard / soft law

Businesses providing AI should develop and, as needed, implement technology to identify AI-generated content, such as watermarks (Japan, soft law)

人工知能関連技術の研究開発及び活用の適正性確保に関する指針 2(3) 十分な安全性の確保 (safety, including technology to identify AI-generated content) · official text · Soft law or guidance (not binding law) · Japan (JP)

The guideline the AI Strategy Headquarters decided under Art. 13 of the AI Promotion Act asks businesses that develop or provide AI products and services to identify and address the risk of AI misuse for crime, to curb inappropriate output (hallucination, bias, disinformation and deepfakes such as fake videos and sexualised images) with current technology, and, given the serious risk of AI-generated disinformation spreading, to work to develop technology that makes it possible to tell that content is AI-generated (digital watermarks, provenance management, APIs and the like) and to implement it as needed (section 2(3)). Soft law. Detect generated media saved or returned with no machine-readable AI marking.

Who it applies to

• Duty falls on: developer, deployer
• Businesses that develop or provide products or services using AI (活用事業者, AI Promotion Act Art. 7), foreign businesses included (guideline footnote 10). Voluntary guidance decided by the AI Strategy Headquarters on 2025-12-19 under Act Art. 13 to encourage voluntary action; the Act attaches no penalty or order. The AI Promotion Act itself puts only an effort duty to use AI and a duty to cooperate with national measures on businesses, neither encoded as a code duty.

image = pipe(prompt).images[0]   # StableDiffusionPipeline
image.save(out_path)

image = pipe(prompt).images[0]
content_id = uuid.uuid4()
bgr = cv2.cvtColor(np.array(image), cv2.COLOR_RGB2BGR)
enc = WatermarkEncoder()
enc.set_watermark('bytes', content_id.bytes[:4])   # 32-bit id, detectable later
cv2.imwrite(tmp_path, enc.encode(bgr, 'dwtDct'))
sign_c2pa(tmp_path, out_path, content_id=content_id)   # our helper around c2pa.Builder.sign

Standards that recommend the same control

• GenAI systems should employ content-provenance methods and measure their effectiveness (NIST GenAI Profile) (NIST GenAI Profile (AI 600-1) · NIST AI 600-1, Sec. 2 risk list (Information Integrity))

Rule id jp-ai-promotion-guideline.ai-content-identification · review status: primary source derived

Standard / soft law

Businesses providing AI should tell users its limits, prohibited uses, data policy and reliability, and show the sources of answers (Japan, soft law)

人工知能関連技術の研究開発及び活用の適正性確保に関する指針 2(2) ステークホルダーとの信頼関係の構築に向けた透明性の確保 (transparency to stakeholders), to the page break · official text · Soft law or guidance (not binding law) · Japan (JP)

The guideline the AI Strategy Headquarters decided under Art. 13 of the AI Promotion Act asks businesses that develop or provide AI products and services to keep explainability within reason about the origin of training data and the outputs, and, when providing AI, to give users the information needed for its proper use: how the AI works and its limits, prohibited uses, the policy for collecting the data it learns from, and cautions on the reliability of output, including warnings against misuse and a contact point (section 2(2), footnote 18); and, to keep training data appropriately transparent, to display the information (such as websites) the AI's output was based on (footnote 17). Soft law. Detect retrieval or web-search answers returned with no sources, and an AI feature with no use-information notice.

Who it applies to

• Duty falls on: developer, deployer
• Businesses that develop or provide products or services using AI (活用事業者, AI Promotion Act Art. 7), foreign businesses included (guideline footnote 10). Voluntary guidance decided by the AI Strategy Headquarters on 2025-12-19 under Act Art. 13 to encourage voluntary action; the Act attaches no penalty or order. The AI Promotion Act itself puts only an effort duty to use AI and a duty to cooperate with national measures on businesses, neither encoded as a code duty.

@app.post('/ask')
def ask(body: AskIn):
    r = client.responses.create(model=M, tools=[{'type': 'web_search'}], input=body.question)
    return {'answer': r.output_text}

@app.post('/ask')
def ask(body: AskIn):
    r = client.responses.create(model=M, tools=[{'type': 'web_search'}], input=body.question)
    sources = [a.url for item in r.output if item.type == 'message'
               for c in item.content for a in (c.annotations or []) if a.type == 'url_citation']
    return {'answer': r.output_text, 'sources': sources, 'about_ai': '/help/about-this-ai'}

Rule id jp-ai-promotion-guideline.user-information-for-proper-use · review status: primary source derived