欧盟将高风险人工智能规则推迟到2027年至2028年,禁止非协商的深层假冒,并通过新的指导方针帮助中小企业. EU delays high-risk AI rules to 2027–2028, bans non-consensual deepfakes, and aids SMEs with new guidelines.

欧洲议会委员会已对人工智能法进行了修正,将高风险的人工智能系统的关键规则推迟到分别2027年12月和2028年8月,以提高法律清晰度. European Parliament committees have advanced amendments to the AI Act, delaying key rules for high-risk AI systems to December 2027 and August 2028, respectively, to improve legal clarity. 禁止非自愿的人工智能产生的亲密内容得到支持, A ban on non-consensual AI-generated intimate content is backed, with exceptions for safeguarded uses. 提供者必须在2026年11月之前满足水印要求. Providers must meet watermarking requirements by November 2026. 这些变化旨在让中小企业和初创公司更容易遵守,简化法规并减少与特定部门法律的重叠. The changes, aimed at easing compliance for SMEs and startups, simplify regulations and reduce overlap with sector-specific laws. 通过了101票,9票的妥协将于3月26日在议会全体投票中进行,随后与理事会谈判. The compromise, approved 101 to 9, heads to a full Parliament vote on March 26, followed by negotiations with the Council. 坦佩雷大学的新指南将于2026年4月18日发布,为中小企业提供了操作规则的实用工具. A new Tampere University guide, publishing April 18, 2026, offers SMEs practical tools to navigate the rules.