AI News

1. OpenAI Cuts Off Cursor's Model Access After SpaceX Acquisition

OpenAI. OpenAI revoked AI coding tool Cursor's access to its models following SpaceX's completed $60 billion acquisition of Cursor maker Anysphere, reportedly citing Elon Musk's history of breaking contracts. The move forces Cursor to lean on rival or in-house models and marks a new front in the feud between OpenAI and Musk's AI ventures. It underscores platform risk for developer tools that depend on a single frontier lab's API. Source

2. Sony Music and Warner Sue Anthropic Over Alleged IP Theft

Sony. Sony Music and Warner filed suit against Anthropic, alleging a "brazen campaign" of intellectual property theft tied to how the company trained its models on copyrighted material. The case adds major music labels to the growing roster of rights holders taking frontier AI labs to court over training data. The outcome could shape licensing norms and liability for generative AI trained on copyrighted works. Source

3. DeepSeek Nears $7.4B Raise at $74B Valuation Ahead of 2027 IPO

DeepSeek. Chinese AI lab DeepSeek is closing a roughly $7.4 billion (50 billion yuan) round at a $74 billion pre-money valuation, with existing backers Monolith, Shixiang Capital, and CATL joined by new investors including CPE and Legend Capital. The company plans to file for an IPO as early as end of 2026, targeting a 2027 debut on Shanghai's Star Market. The raise signals DeepSeek's shift from a founder-financed research project to a heavily capitalized commercial player scaling compute. Source

4. LAION Releases Open Video Dataset With 10 Million Hours of Footage

LAION. Nonprofit LAION released its Big Video Dataset, built from 80 million downloaded videos totaling roughly 10 million hours, yielding about 55 million extracted clips and 300 million still images for AI research. The open corpus targets training and benchmarking of video understanding and generation models, a data-scarce area relative to text and images. It gives academic and open-source teams a large, freely available foundation for multimodal work. Source

5. Report Says Always-On, Self-Starting Agents May Be OpenAI's Next Play

OpenAI. Reporting points to OpenAI building a "Persistent Mode" for its Codex agent that would keep working proactively until told to stop, rather than acting only on discrete prompts. The shift toward always-on, self-initiating agents would move autonomous coding and task execution closer to continuous background operation. It flags a direction practitioners may soon need to design workflows and guardrails around. Source

6. Nvidia's AI Advantage Is Moving Beyond the GPU

Nvidia. A TechCrunch analysis argues Nvidia's competitive moat is shifting from raw GPU performance to system-level advantages in networking, interconnects, and full-rack integration that are harder for rivals to replicate. The framing matters as custom silicon from hyperscalers and Chinese accelerators chip away at the chip-only story. It suggests buyers evaluating AI infrastructure should weigh whole-system economics, not just per-chip specs. Source

7. AI-Generated Video Is Already Displacing Actors and Livestreamers in China

China. A report describes AI-generated video already replacing actors and livestreamers across China's entertainment and e-commerce sectors, with synthetic presenters and performers cutting production costs. The trend offers an early, concrete look at generative video's labor impact in a large market. It signals to practitioners how quickly video models are moving from demos into commercial displacement of on-camera work. Source