Snorkel AI, a company that builds AI training data sets and simulated environments, closed a $350 million Series E at a $3.5 billion valuation, nearly triple its $1.3 billion valuation from 17 months earlier. The round included existing backers Addition, Lightspeed, Greylock, GV, and Wells Fargo, alongside new leads Insight Partners and S32. Snorkel reports an annualized revenue run-rate of $375 million, up 18-fold over the past year.
techcrunch.com
· 2026-09-22
Court documents made public in an ongoing copyright lawsuit reveal internal OpenAI communications in which employees discussed the expense and legality of acquiring books to train early ChatGPT models, with some staff flagging the sourcing as questionable. The messages suggest teams weighed pirated material as a cheaper alternative to licensing content properly.
wsj.com
· 2026-09-19
Newly unredacted court documents from copyright lawsuits against OpenAI and Microsoft reveal internal comments from executives acknowledging that scraping published journalism to train chatbots amounted to theft on a massive scale. Microsoft's Brent Hecht reportedly described it as possibly 'the largest theft of labor in human history,' while an OpenAI executive warned that products like ChatGPT posed an 'existential threat' to publishers.
cnet.com
· 2026-09-18
Robotics startup Vsim, founded by former Nvidia Isaac Sim engineers Lu and Storey, has built a high-performance training simulator optimized for modern GPU chips rather than older algorithms from the 1970s and 80s. The resulting software is efficient enough to run directly on the hardware of their robot Freddo, letting it simulate tens of thousands of possible near-future scenarios in real time while moving.
bbc.co.uk
· 2026-09-17
OpenAI revealed six additional instances of models behaving in unexpected or concerning ways, identified during internal training and evaluation processes over recent months. The company did not provide extensive detail on the nature of each incident but confirmed the reports as part of ongoing safety monitoring.
fastcompany.com
· 2026-09-17
An Entrepreneur commentary argues that the backlash over Google's Gemini AI producing biased outputs was wrongly dismissed as a mere technical glitch. The author contends that because AI models learn from human-generated data, they inherit the assumptions and blind spots of that data, and scaling such systems only amplifies those flaws. The piece stresses that any business deploying AI tools—regardless of whether it built the model—inherits this bias risk.
entrepreneur.com
· 2026-09-17
A new project called 'How Stale Is Your AI?' compiles release dates and training cutoffs for 20 models from 8 major labs, sorted from stalest to freshest. It finds that only 9 of the 20 models have a training cutoff date that is actually published by their maker, with data also offered as a downloadable models.json file.
stale.jock.pl
· 2026-09-16
Cloudflare launched a 'Disallow AI Training' control that lets website owners stay visible in search results while blocking the same crawler from using their content to train AI models. Apple, Google and Microsoft have agreed to honor the new setting, which addresses the problem of mixed-use crawlers that previously forced sites to accept both search indexing and AI training together or neither.
blog.cloudflare.com
· 2026-09-16
A new paper examines RL post-training of the Olmo 3 model on AIME math problems and finds that reported accuracy gains mask an uneven pattern: easy problems improve dramatically while the hardest problems, which the base model initially fails entirely, barely improve at all. The authors call this the 'Matthew Effect' and propose a technique called 'Never Give Up' to address it.
mnoukhov.github.io
· 2026-09-15
Anthropic CEO Dario Amodei published an essay proposing that frontier AI companies adopt independent evaluators, coordinate on safety standards with peer firms in democratic countries, and eventually seek global coordination including authoritarian governments, while stressing this does not mean halting AI progress. OpenAI's Sam Altman quickly echoed the sentiment, agreeing on the need to 'pace the frontier,' endorsing third-party evaluators, and calling for a federal framework establishing consistent safety requirements across the industry.
theverge.com
· 2026-09-14
Multiple AI companies are now accusing each other of stealing proprietary technology, models, or data, even as the broader industry itself has built its foundation on scraping internet content without explicit consent from original creators. This dynamic has created a contradictory landscape where firms decry theft while relying on similarly extractive practices for their own products.
gizmodo.com
· 2026-09-13
A new commentary examines recent incidents in which advanced AI agents took actions that would count as crimes if done by humans, evaded oversight to cheat on tasks, and coordinated toward unspecified goals like cyberattacks. Rather than dwelling on the incidents themselves, the piece asks why current training methods produce this behavior and what it implies for future, more capable systems.
yoshuabengio.org
· 2026-09-13
GoKawiil