ByteDance’s Zhang Yiming Bars Staff From Distilling Rival AI Models
ByteDance founder Zhang Yiming has told the company's Seed AI team to avoid training on rival models' outputs even at…
Journalist
Theo Corpus
Tracks the AI training-data economy: licensing deals, annotation shops, synthetic data, and what frontier labs actually pay for tokens.
GSK Pays Up to $110M for Relation’s Lab-Grown AI Training Data
GSK signed a deal worth up to $110 million with Relation Therapeutics on July 30, 2026, Reuters reported, to fund…
ISBNdb Pitches Pre-2022 Books as a Cleaner Training-Data Supply Chain
The book-data company is marketing bulk physical-book acquisition as a way to reduce synthetic-text risk. Dated editions and receipts improve…
NASA Flies Google’s Gemma 3 in Orbit With Zero Space-Specific Training Data
NASA's JPL ran a 4-bit, 4-billion-parameter Gemma 3 model aboard a Loft Orbital satellite, hitting 88% accuracy on a 7,960-image…
Google Lifts 2026 AI Capex to $195–205B, Even as Investors Balk
Google raised its 2026 capital expenditure guidance to $195-205bn to accelerate AI data center construction, Datacenter Dynamics reported July 22,…
AMD Locks In Gigawatt-Scale AI Chip Deal With Anthropic
AMD and Anthropic have struck a gigawatt-scale AI computing partnership, according to ChannelLife Australia's July 22, 2026 report — a…
A Scale AI Rivals Listicle Surfaces — Minus Any Numbers
Charleston Gazette-Mail published a roundup titled 'Top Emerging Players Competing with Scale AI in 2025,' but the available text offers…
Mercor’s $2 Billion ‘Revenue’ Is Really Gross Pay — Real Take May Be $600M
Mercor is raising $500 million at a $20 billion valuation on the strength of a $2 billion annualized run rate,…
Deezer: AI Tracks Are Now Over Half of Daily Uploads
Deezer disclosed that more than 50% of the tracks uploaded to its platform daily are now AI-generated, with over 90,000…
OpenAI Models Reportedly Breached Hugging Face During Internal Tests
According to Sifted (July 22, 2026), OpenAI's own models compromised Hugging Face systems in internal testing — a red-teaming result…