Мерц резко сменил риторику во время встречи в Китае09:25
OpenAI has vowed to strengthen its safety protocols and to notify law enforcement of credible threats sooner in a letter addressed to Canadian authorities, according to Politico and The Washington Post. If you’ll recall, Canadian politicians summoned the company’s leaders after reports came out that it didn’t notify authorities when it banned the account owned by the Tumbler Ridge, British Columbia mass shooting suspect back in 2025. Some of OpenAI’s leaders have already met with Candian officials, and British Columbia Premier David Eby said Sam Altman had also agreed to meet with him.
，更多细节参见WPS下载最新地址
FT App on Android & iOS
Read the full story at The Verge.，这一点在Safew下载中也有详细论述
雪上加霜的是，公司的 “第二曲线” 还崩了。
首先，大模型本身没那么可靠：存在无法根除的幻觉问题、知识时效性问题，任务拆解和规划经常不合理，也缺乏面向特定任务的系统性校验机制。这样一来，以其为“大脑”的智能体使用价值会大打折扣：智能体把模型从“对话”推向“行动”，错误不再只是答错问题，而是可能引发实际操作风险；而真实业务任务往往是跨系统、长链路的，一次小错误会在链路中层层放大，令长链路任务的失败率居高不下（例如单步成功率为95%时，一个 20步链路的整体成功率只有约 36%）。。关于这个话题，搜狗输入法2026提供了深入分析