内容来源：https://aibusiness.com/ai-policy/new-york-signs-ai-safety-legislation

内容总结：

纽约州率先立法规范高级人工智能 安全监管与联邦政策形成角力

在生成式人工智能应用浪潮中，如何确保其安全发展已成为全球监管焦点。近日，美国纽约州在这一领域迈出关键一步，其立法行动也与联邦层面的政策取向形成了微妙张力。

12月19日，纽约州州长凯西·霍楚签署了名为《负责任人工智能与安全教育法案》的立法。该法案旨在为最先进的人工智能模型设定安全规则，并计划于2027年1月1日正式生效。法案核心内容要求，年收入超过5亿美元、开发人工智能系统的公司必须制定并公布详细的安全协议，并在发生安全事件后72小时内向州政府报告。违规企业将面临最高100万美元的初犯罚款，屡犯者罚金可高达300万美元。

值得注意的是，此项立法的时间点颇具深意。它是在前总统特朗普签署行政令，旨在将人工智能监管权收归联邦、以避免其所谓的“各州过度监管”后不久迅速获批的。霍楚州长对此直言，该法律借鉴了加州今年早些时候通过的框架，旨在联邦政府监管滞后的情况下，于领先的科技州之间建立统一的基准，以实施保护公众的常识性法规。

尽管经历了科技公司数月的游说，法案中部分最初提议的严厉条款（如禁止公司发布未通过安全测试的模型）已被削弱，罚款金额也有所降低，但法案的主要发起人、州议员亚历克斯·博尔斯仍对结果表示满意。他强调，纽约州挫败了“人工智能寡头”扼杀该法案的最后尝试，从而为人工智能安全立法树立了底线，同时也抵制了特朗普及其捐助者试图通过行政令营造人工智能“蛮荒西部”环境的行动。

纽约州的此次立法，标志着美国在人工智能治理上已呈现出联邦与州政府并行探索、甚至存在博弈的复杂局面，各科技重镇正试图在联邦法规缺位时自行设定规则。随着2027年生效日期的临近，该法案的实际执行效果及其与未来联邦政策的互动，将持续受到业界与公众的关注。

中文翻译：

由谷歌云赞助
选择您的首批生成式人工智能应用场景
要开始应用生成式人工智能，首先应关注能够提升人类信息交互体验的领域。

纽约州州长凯西·霍楚于12月19日签署了一项法案，旨在为最先进的人工智能模型制定安全规则。
《负责任人工智能与安全教育法案》的出台具有多重重要意义，其时机尤为关键——该法案获批之际，正值前总统唐纳德·特朗普签署行政令，试图将人工智能监管权收归联邦政府，以避免其所谓的"各州过度监管"。

几乎立即有预测指出，各州将发起抵制该行政令的行动。而率先迎接挑战的正是纽约州——霍楚批准了这项于今年6月获州议会通过的法案。
霍楚解释道："这项法律借鉴了加州近期通过的监管框架，在联邦政府监管滞后的情况下，为美国领先的科技州建立了统一基准。联邦层面至今未能实施保护公众的常识性法规。"
加州的相关法案于今年早些时候通过，早于特朗普行政令的颁布。

纽约州法案的核心要求包括：自2027年1月1日起，任何年收入超过5亿美元的人工智能系统开发公司，都必须制定并公布其安全协议的具体内容。
企业还须在72小时内向州政府报告安全事故。同时，纽约州将在金融服务部内设立监督办公室，负责评估开发企业合规情况，该部门将每年发布评估报告。
违规企业将面临最高100万美元的首次处罚，后续违规罚款可高达300万美元。

在科技公司长达数月的游说后，霍楚最终批准了该法案。据报道，部分原定措施已被搁置，包括禁止企业发布未通过安全测试的模型。最初提议的更高额罚款标准也有所降低。
尽管如此，法案发起人、纽约州议员亚历克斯·博尔斯对最终结果表示满意。
博尔斯表示："在纽约州，我们挫败了人工智能寡头的最后挣扎，成功守住了这项法案，从而为人工智能安全立法树立了底线标准。我们也击退了特朗普及其资助者试图通过行政令阻挠《负责任人工智能与安全教育法案》实施的企图——那种做法只会纵容人工智能领域沦为'蛮荒西部'。"

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
New York state Governor Kathy Hochul signed legislation Dec. 19 designed to set safety rules for the most advanced AI models.
The Responsible Artificial Intelligence and Safety Education (RAISE) Act is notable for several reasons, not the least of which is timing as its approval comes just days after President Donald Trump signed an executive order to bring AI under federal control rather than be subjected to what he considers "excessive state regulation."
Almost immediately it was predicted that the states would mount a fight to resist the executive order. And first to pick up the challenge is New York, with Hochul greenlighting rules that were originally approved by state lawmakers in June.
Hochul explained: "The law builds on California's recently adopted framework, creating a unified benchmark among the country's leading tech states as the federal government lags behind, failing to implement common sense regulations that protect the public."
California's bill was passed earlier this year, before the Trump executive order was enacted.
Among the key requirements of the New York bill, which will be enforced starting Jan. 1 2027, is that any company with more than $500 million in revenue that develops an AI system must create and publish details of their safety protocols.
They must also report incidents to the state within 72 hours, while New York itself is creating an oversight office within the Department of Financial Services to assess developers. The department will issue annual reports.
Companies that fail to comply with the regulations could be fined up to $1 million for first-time offenses and up to $3 million for subsequent violations.
Hochul's approval was granted after months of lobbying from tech companies, with reports that some measures — including banning companies from releasing models that do not pass safety tests — have fallen by the wayside. Steeper fines originally proposed have also been reduced.
Nevertheless, bill sponsor and New York assembly member Alex Bores declared himself satisfied with the outcome.
Bores said: " In New York, we defeated last-ditch attempts from AI oligarchs to wipe out this bill and, by doing so, raised the floor for what AI safety legislation can look like. And we defeated Trump's — and his donors' — attempt to stop RAISE through executive action greenlighting a Wild West for AI.