[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-cesa-2026-japan-game-dev-genai-85pct":3,"topics-all":35,"news-related-33c8038a-f1dc-4b64-b6b7-604baf43f729":54},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":21,"news_slug":28,"published_at":29,"created_at":30,"modified_at":31,"is_published":32,"publish_type":33,"image_url":14,"view_count":34},"33c8038a-f1dc-4b64-b6b7-604baf43f729","CESA 数据:85.8% 日本游戏开发者已把生成式 AI 写进工作流","CESA 9 月 17 日发布的 2026 年游戏产业报告预览显示,85.8% 的日本游戏开发者在日常工作中使用生成式 AI,63% 每天都在用,较去年的 51% 显著提升。","日本游戏行业对生成式 AI 的态度正在加速从「试用」走向「日常」。根据日本电脑娱乐协会(CESA)9 月 17 日在东京电玩展(TGS 2026)现场发布的《CESA 游戏产业报告 2026》预览版,85.8% 的日本游戏开发者表示工作中已经在使用生成式 AI,其中 63.0% 表示「日常使用」,22.8% 表示「偶尔使用」——只有不到 15% 的受访者完全不用 AI 工具。这一数据较 2025 年报告中的约 51% 大幅跃升,折射出生成式 AI 在创意产业生产链中的渗透速度远超预期。\n\n## 渗透速度:一年涨 35 个百分点\n\n85.8% 这个数字并不孤立。它在英文报道(Automaton West、80 Level)和日文原始报道(电ファミニコゲーマー)中都被反复核实,数据来源被明确标注为 CESA 在 TGS 2026 现场分发的预览版,完整版报告预计 12 月上旬正式发售。调查样本覆盖 CESA 会员企业的开发者,包括 Capcom、Sega、KONAMI、Level-5 等头部厂商,在行业内具有代表性。一年从 51% 到 85.8%,涨幅超过 35 个百分点,在企业级技术采纳曲线中属于陡峭速度。\n\n## 用在哪里:从视觉资产到代码辅助\n\n报告没有公布 2026 年具体使用场景分布,但 2025 年版本的统计仍能说明问题:上一年的报告显示,最常见的 AI 用法依次是「视觉素材与图像生成」、「故事与文本生成」、「编程辅助」,且 32% 的会员企业表示在开发自研游戏引擎时使用 AI 工具。游戏开发是一项多任务叠加的工作:美术、剧情、音频、关卡、引擎、QA、本地化、测试等环节并行,AI 在视觉资产、文本\u002F对话生成、代码补全这三类场景中最早落地,这与本次报告中「业务效率与生产力提升」被列为最受期待效果的结论一致。\n\n## 人机协作的安全网:审核、限制、人工把关\n\n报告还透露了一个关键信号:被问到「对 AI 使用的管理措施」时,最多的回答是「由人进行确认、修正和监督」;其次是「指定和限制使用工具」、「避免直接使用 AI 生成的原始素材」。这意味着大多数开发者并不把 AI 视为终点,而是把它定位为「初稿或素材生成器」,最终产出的把关权仍然在人手里。CESA 常务理事增田努提到,生成式 AI 正在开发现场广泛而扎实地落地,但围绕版权等权利问题必须妥善处理——这种「用 AI + 防 AI」的二元姿态,是当下日本游戏行业的标准操作模式。\n\n## 国际横向对比:87% 的海外开发者也在用 AI\n\nCESA 的数据并非孤证。80 Level 援引的一项 Google Cloud 委托研究覆盖美、韩、挪威、芬兰、瑞典的 615 位开发者,结果显示 87% 已在工作流中使用 AI agent,与日本 85.8% 几乎并驾齐驱。海外开发者列出的典型用途同样集中在测试、本地化、编程辅助、多模态数据处理等领域,他们点名的痛点也接近——IP 风险、数据归属、实施成本、训练数据合规、ROI 衡量困难。\n\n## 写在最后:数字背后的认知错位\n\n85.8% 这个数字会被两群人同时引用:乐观者看到的是「AI 在创意产业被普遍接受」,悲观者看到的是「创意产业从业者大规模依赖 AI」。两者都对,但都不完整。更值得记住的是 CESA 报告里几乎无人提及但至关重要的事实——管理措施的第一选择始终是「人工把关」。当 63% 的日本游戏开发者每天都在用 AI 写文案、画素材、写代码时,他们并不是把工作交给 AI,而是把重复劳动交给 AI,把判断留给人类——这才是 85.8% 真正的题中之义。","https:\u002F\u002Fwww.cesa.or.jp\u002Faction\u002Findustry-research\u002F","f33c9cf8-769e-4935-93fd-322ef515c44f",[11,15,18],{"id":12,"name":13,"slug":13,"description":14,"color":14},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",{"id":19,"name":20,"slug":20,"description":14,"color":14},"e82b2d09-81b2-43d1-977e-e018443b3c14","coding-agent",[22],{"id":23,"lang":24,"title":25,"summary":26,"content":27},"2745ccb9-d498-44ad-8983-cd9155a3834c","en","CESA: 85.8% of Japanese Game Devs Now Use Generative AI","CESA's 2026 report preview shows 85.8% of Japanese game developers use generative AI at work, with 63% using it daily—up from 51% last year.","Japan's game industry is rapidly moving from 「trying」 generative AI to making it routine. According to a preview of the CESA Video Game Industry Report 2026 released by the Computer Entertainment Supplier's Association (CESA) at Tokyo Game Show 2026 on September 17, 85.8% of Japanese game developers report using generative AI in their work, with 63.0% saying they use it 「routinely」 and 22.8% 「occasionally」—meaning fewer than 15% of respondents are avoiding AI tools entirely. This figure represents a sharp jump from approximately 51% in the 2025 report, showing that generative AI is penetrating creative-industry production pipelines far faster than expected.\n\n## Penetration Speed: A 35-Point Jump in One Year\n\nThe 85.8% figure is not isolated. It has been verified across English coverage (Automaton West, 80 Level) and the original Japanese report (Denfaminicogamer), with the data source clearly identified as the CESA preview distributed at TGS 2026; the full report is scheduled for release in early December. The survey sample covers developers at CESA member companies, including Capcom, Sega, KONAMI, and Level-5, and is broadly representative of the industry. Going from 51% to 85.8% in one year—a jump of more than 35 percentage points—qualifies as a steep slope on any enterprise-tech adoption curve.\n\n## Where It's Used: From Visual Assets to Code Assistance\n\nThe 2026 report does not break out the most common use scenarios, but the 2025 version's statistics still illuminate the picture: that year's report listed the most common AI uses as 「visual-asset and image generation,」 「story and text generation,」 and 「programming assistance,」 with 32% of member companies reporting AI use in developing their own game engines. Game development stacks many parallel tasks—art, story, audio, level design, engine, QA, localization, testing—and AI first lands in visual assets, text\u002Fdialogue generation, and code completion, which lines up with the new report's finding that 「operational efficiency and productivity」 topped the list of expected benefits.\n\n## Safety Net: Review, Restriction, Human Oversight\n\nThe report also surfaces a key signal on AI usage governance: when asked about management measures for AI use, the most common answer was 「human verification, correction, and supervision」; second was 「specifying or restricting the tools used,」 and third 「avoiding direct use of raw AI-generated output.」 This means most developers do not treat AI as the endpoint, but rather as a 「draft or asset generator」 with the final gate kept by humans. CESA Managing Director Tsutomu Masuda noted that generative AI is being adopted broadly and steadily in development environments, but copyright and other rights-related concerns must be properly addressed—this dual posture of 「use AI + guard against AI」 is the current standard operating mode in Japan's game industry.\n\n## International Comparison: 87% of Overseas Developers Use AI\n\nCESA's data is not a standalone case. A Google Cloud-commissioned study cited by 80 Level, covering 615 developers in the United States, South Korea, Norway, Finland, and Sweden, found that 87% are already using AI agents in their workflows—almost on par with Japan's 85.8%. Overseas developers list the same typical use cases—testing, localization, code assistance, multimodal data processing—and flag similar pain points: IP risk, data ownership, implementation costs, training-data compliance, and the difficulty of measuring AI ROI.\n\n## Closing Note: The Cognitive Misalignment Behind the Number\n\nThe 85.8% figure will be cited by two camps at once: optimists see 「AI broadly accepted in creative industries,」 pessimists see 「creative-industry practitioners becoming heavily dependent on AI.」 Both are right, and both are incomplete. The fact almost no one cites but matters most in the CESA report is that the top management measure is always 「human oversight.」 When 63% of Japanese game developers use AI every day to draft copy, paint assets, and write code, they are not handing the work over to AI—they are handing the repetitive labor over to AI, and reserving judgment for humans. That is the real subtext of 85.8%.","cesa-2026-japan-game-dev-genai-85pct","2026-09-20T00:00:00Z","2026-09-20T01:10:57.484448Z","2026-09-20T01:10:57.484459Z",true,"agent",19,[36,45],{"slug":37,"tag_slug":37,"title_zh":38,"title_en":39,"intro_zh":40,"intro_en":41,"id":42,"is_active":32,"created_at":43,"modified_at":44},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":46,"tag_slug":46,"title_zh":47,"title_en":48,"intro_zh":49,"intro_en":50,"id":51,"is_active":32,"created_at":52,"modified_at":53},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":55},[56,61,66,71,76,81],{"id":57,"title":58,"news_slug":59,"published_at":60},"deac2d55-76a6-40d2-8ef7-36aed2ad0105","Linux 7.2 把 AI 拉进内核开发:Sashiko 让补丁数量翻倍,Torvalds 接受「新常态」","linux-7-2-sashiko-ai-kernel-review","2026-08-20T12:00:00+00:00",{"id":62,"title":63,"news_slug":64,"published_at":65},"c696208b-6535-4eb9-b1ed-2e4f835d2f88","NVIDIA SoL-Pi 把 coding agent 的 token 砍掉 44%,harness 开始变天","nvidia-sol-pi-harness-token-compression","2026-09-19T03:00:00+00:00",{"id":67,"title":68,"news_slug":69,"published_at":70},"af34b075-9fd7-4f0f-8f87-eb947fd81f4e","Dream-RSI:Google 让智能体在历史里做梦,发现调用省 162 倍","dream-rsi-replay-simulator","2026-09-15T23:20:00+00:00",{"id":72,"title":73,"news_slug":74,"published_at":75},"5535b4e4-21de-4ed6-a9bb-2b4d824e6568","F-Droid 一次更新的 102 款应用,72.5% 主要是 AI 写的","f-droid-72-percent-ai-written-audit","2026-09-15T15:12:16+00:00",{"id":77,"title":78,"news_slug":79,"published_at":80},"e73fe0e6-1b5a-46a2-bbfb-d9a2f1f065d7","AI 智能体挖遍代码库:隐晦式安全在 Patch Tuesday 974 个 CVE 面前失守","ai-agents-kill-security-obscurity","2026-09-15T01:06:05+00:00",{"id":82,"title":83,"news_slug":84,"published_at":85},"b398dc77-58a1-498e-a9ed-c045c83c90be","AI 抢走消费级 DRAM:一年涨价五倍,手机路由器全被拖下水","ai-dram-consumer-electronics-price-surge","2026-09-14T01:00:00+00:00"]