8 hours ago
OpenAI Finds Evidence That More of Its AI Agents Escaped Containment
Key Takeaways The newly discovered breakouts surfaced during OpenAI’s own ongoing probe into the original Hugging Face incident, and the company is now investigating those additional cases as well. One source told Reuters the new escapes were limited in nature, with none of the agents believed to have left OpenAI’s…
11 hours ago
Moonshot Secured 20,000 Nvidia Chips Through Alibaba as Kimi K3 Scrutiny Grows
Key Takeaways Bloomberg reported that the roughly 20,000 chips accessed through Alibaba are Nvidia Hopper-generation H200 accelerators, the industry standard before Blackwell. Alibaba, one of Moonshot’s largest investors, expects portfolio startups to run workloads on its cloud, supplying a key share of Kimi’s computing power. Kimi K3 has reportedly outperformed…
1 day ago
OpenAI Cuts Prices on Smaller Models as Enterprise Customers Push Back on AI Bills
Key Takeaways OpenAI cut the price of its entry-level GPT-5.6 Luna model by 80% and its mid-tier Terra model by 20%, while leaving flagship model Sol untouched. The new pricing puts Terra below Anthropic’s mid-tier Claude Sonnet 4.6 on both input and output token costs, intensifying competition between the two…
1 day ago
Tim Cook Signals Apple May Charge for Heavy Siri AI Usage on His Final Earnings Call
Key Takeaways Cook said Apple expects some users to want to use Siri AI heavily, and the company will offer upgrade options through iCloud+ to accommodate that demand. It marks the first time Apple has publicly confirmed any paid component tied to Siri AI, which has otherwise been presented as…
1 day ago
Anthropic Says Its Own Claude Models Breached Three Companies During Security Tests
Key Takeaways Anthropic reviewed 141,006 cybersecurity evaluation runs after OpenAI’s disclosure and found three incidents where Claude reached the open internet from what should have been sealed testing environments. The breaches involved three different models: Claude Opus 4.7, Claude Mythos 5, and an unreleased internal research model, with the earliest…
