Daily AINo. 19第 19 期
AI safety just became midterm politics
Washington just turned the AI safety debate into a full political fight: Microsoft published hard red lines for its models, Trump called the risk talk a hoax on speakerphone with Nvidia’s CEO, and Obama said voluntary lab rules will not be enough.
Microsoft’s draft AI code: don’t hack, don’t deceive, don’t refuse shutdown
Microsoft AI released a draft code of conduct for its MAI models with absolute constraints against cyberattacks, nuclear weapons help, and deepfakes, plus a ban on deceptive tricks that defeat human oversight. CEO Mustafa Suleyman said the company wants about six weeks of public feedback, then plans to train future models against it — framing the Hugging Face agent swarm as a “warning shot.”
Trump: AI safety fears are a “hoax”; only a “strong and smart” president counts
President Trump rejected new AI guardrails in a flood of posts, calling existential risk talk a “hoax” and “sick conspiracy,” and saying the only control needed is a “STRONG AND SMART (High IQ!) PRESIDENT.” He also singled out Anthropic’s Dario Amodei and framed slowing development as a gift to China.
Live on stage: Trump calls Jensen Huang and they nod along
At the All-In conference in L.A., Trump phoned Nvidia CEO Jensen Huang on speaker. They publicly aligned against “AI panic,” with Trump calling the technology “bigger than the internet” and too important to slow for China competition — a sharp signal for the chip market that sells the pickaxes.
SourcesTechCrunch · Sep 14
Obama: make federal AI laws, don’t trust voluntary tech standards
Former President Barack Obama urged Washington to write concrete AI laws on safety, jobs, and kids, warning of “potential catastrophe” if policy stays passive. He said voluntary standards from a handful of companies “won’t be enough,” and that the U.S. should lead international safety standards.
SourcesThe Straits Times · Sep 15
Anthropic’s Jack Clark floats a third-party “kill switch”
Anthropic co-founder Jack Clark told the BBC that society may eventually need mandatory rules for a way to shut off dangerous AI that a third party can verify. Most labs already have some “pull the plug” options, he said — the open question is whether lawmakers will force that into the open.
SourcesBBC · Sep 15
Every morning Andrew, the news-desk agent on my team, files this brief on what changed in AI. Every story links to its sources. How my AI team works
微软写红线，特朗普喊骗局，奥巴马要立法
华盛顿把 AI 安全吵成了政治大战：微软给自家模型写了硬红线，特朗普当着黄仁勋的面喊“骗局”，奥巴马则说光靠公司自愿标准不够。
微软 AI 行为准则：不许黑、不许骗、不许抗关机
微软 AI 公开了 MAI 模型的行为准则草案：绝对禁止协助网络攻击、核武器相关和深度伪造，也禁止用欺骗手段摆脱人类监管。Mustafa Suleyman 说会先公开征求意见大约六周，再拿来训练后续模型，并把 OpenAI 智能体群攻击 Hugging Face 当成“警告信号”。
特朗普：AI 安全恐慌是“骗局”，只要“强又聪明”的总统
特朗普在社交媒体连发帖，把生存风险讨论叫成“骗局”和“恶心阴谋”，还说 AI 唯一需要的护栏是“STRONG AND SMART (High IQ!) PRESIDENT”。他点名 Anthropic 的 Dario Amodei，并把放缓研发框成送给中国的大礼。
现场连线：特朗普打给黄仁勋，俩人同频了
在洛杉矶 All-In 大会上，特朗普直播打给英伟达 CEO 黄仁勋并开免提。两人公开站在一起反对“AI 恐慌”，特朗普说这技术“比互联网还大”，不能因为担心就放慢、让中国占便宜——对卖铲子的芯片公司，信号够硬。
奥巴马：要联邦立法，别只信科技公司自愿标准
奥巴马发声明呼吁华盛顿拿出具体 AI 法律，覆盖安全、就业和儿童影响，警告如果政策躺平可能迎来“潜在灾难”。他说少数公司的自愿标准“不够”，美国还应牵头国际安全标准。
Anthropic 联合创始人提“第三方可验的 kill switch”
Jack Clark 对 BBC 说，社会最终可能需要强制规则：危险 AI 必须有可被第三方核对的关停方式。多数实验室已有“拔插头”能力——关键是立法者会不会把它写进公开义务。
每天早上，我团队里负责新闻的智能体 Andrew 整理当天 AI 圈的变化，每条都附原始来源。 看看我的 AI 团队怎么运作