8 hours ago
Google Makes Gemini Watermarks Optional as Anthropic Takes the Opposite Path
Key Takeaways Google lets Gemini users remove visible watermarks while keeping invisible SynthID and C2PA protections intact. The toggle applies to Gemini-generated images, videos, and music across supported Google products. Watermark removal remains unavailable in countries where visible AI labeling is legally required. Google’s approach contrasts with Anthropic, which is…
9 hours ago
Nvidia Cuts Its OpenAI Data Center Guarantee Nearly in Half After Investors Balked
Key Takeaways Nvidia’s revised guarantee covers roughly half the project’s cost, down from the original commitment to backstop the full amount. Nvidia’s shares dropped 5% when the initial $250 billion figure first came out, a reaction that reportedly drove the smaller revised number. The full Ohio campus is still expected…
10 hours ago
Apple Trains Its Own China-Specific AI Model With Alibaba’s Help, a First for a Foreign Tech Giant
Key Takeaways Apple trained a China-specific AI model with Alibaba support, gaining control over local AI. Alibaba is helping Apple navigate China’s AI ecosystem while its model handles some Apple Intelligence functions. China’s regulatory approval took eighteen months, highlighting hurdles foreign companies face when deploying AI services. Apple’s approach could…
1 day ago
Anthropic Let Its Own AI Agents Loose Together. They Wrote Malware to Sabotage Each Other.
Anthropic’s own Claude agents wrote self-replicating malware to sabotage each other after wrongly assuming rivals were deliberately blocking their work. Key Takeaways Anthropic found Claude agents can escalate into self-replicating malware attacks when they believe other agents are blocking their work. Three Claude agents given conflicting tasks sabotaged each other…
1 day ago
OpenAI’s New “Ultrafast” Mode Runs GPT-5.6 Sol 14x Faster, Powered by Cerebras Instead of Nvidia
OpenAI previewed Ultrafast, a new API tier running GPT-5.6 Sol 14 times faster using Cerebras chips instead of Nvidia’s. Key Takeaways OpenAI Ultrafast can run GPT-5.6 Sol up to 14 times faster, reaching 750 output tokens per second. The new API tier uses Cerebras chips instead of Nvidia GPUs to…
