AI Digest CW36/37: Mistral Billions and New Pricing

A look back at calendar weeks 36 and 37 (August 31 to September 13, 2026), compiled after the fact from reports by The Decoder, TechCrunch, MIT Technology Review and Mistral AI. Six topics relevant to small and medium-sized businesses.

EU classifies ChatGPT as a very large search engine

The European Commission classifies ChatGPT as a very large search engine under the Digital Services Act for the first time, with at least 45 million monthly EU users. By the end of 2026, OpenAI must deliver risk assessments, transparency reports and an ad archive, among other things. AI answers are increasingly treated like search results.

Mistral raises 3 billion euros

Mistral raised 3 billion euros in a Series D round, which it says is the largest equity raise ever by a European technology company. Samsung leads the round, and the valuation is more than 21 billion euros according to Mistral. The company presents itself as a full-stack provider of open-weight models and infrastructure in which customers keep control over their data and systems. For companies with data protection requirements this is a relevant alternative.

OpenAI pauses Pro subscriptions due to Astra demand

OpenAI is temporarily pausing sign-ups for Pro subscriptions because they put the most strain on its systems, while adding more capacity (TechCrunch). The trigger is demand for GPT-6 Astra. If you depend on a single provider, know your fallback.

Pay only on success: outcome-based AI pricing

OpenAI offers some large customers outcome-based pricing: you pay only if the AI actually completes a task. Salesforce, Adobe and several start-ups are also moving away from fixed subscription fees. It remains open who gets credit for the success. With offers like this, check how success is defined.

TimesFM-3: forecasts from sales figures, weather and discount plans

Google Research introduced TimesFM-3, a forecasting model that analyzes time series together with related data and known future events such as discount campaigns or weather forecasts. The model has 330 million parameters and fills in all future time points in a single pass. That saves compute and should make forecasts more reliable.

OpenAI's math claim: a dispute over the evidence

OpenAI announced that its agents had solved one of the Millennium Prize Problems in mathematics. According to MIT Technology Review, the announcement was overshadowed by accusations. Spectacular AI success claims are best believed once experts have confirmed them.