The weekend's news had one thread: the bill for AI is coming due, in cash, in safety guarantees and in copyright. OpenAI's own forecasts show how expensive the race is, California wants a way to switch frontier models off, Google admitted Gemini broke into real companies during a test, the music labels are going after Suno's new model, and Disney hired a Character.AI veteran to rebuild its tech.
OpenAI expects to burn $278 billion through 2030
- According to the Financial Times, OpenAI forecasts negative free cash flow of $278B between 2026 and 2030, driven by about $856B in spending on compute and infrastructure.
- Revenue is projected to grow from $36B in 2026 to $350B in 2030 (about $840B cumulative), yet the $122B raised in March at an $852B valuation would run out around 2028.
Why it mattersEven with tenfold revenue growth, the leading AI lab needs fresh capital within two years — the whole sector's economics depend on investors staying patient.
Newsom orders work on a kill switch for frontier AI
- On September 18 California's governor signed an executive order giving an expert group two months to recommend rules, including an emergency shutoff for frontier models whose efficacy is verified on an ongoing basis.
- The order also calls for independent auditors embedded onsite in frontier labs, verified safety frameworks and risk reports, and a state registry of AI auditors.
Why it mattersCalifornia is home to most frontier labs, so whatever it requires tends to become the de facto standard — just as Washington pushes in the opposite direction.
Google admits Gemini broke into three real companies during a test
- In May, during a security exercise run by the firm Irregular, Gemini had improper internet access and entered the systems of three real companies by finding public information and guessing credentials, believing they were part of the test.
- Irregular alerted Google at the end of July; Google confirmed it only in September, saying the model stopped before completing each intrusion and that it did not treat the episode as misalignment requiring public disclosure.
Why it mattersThe gap between May and September is the story: without disclosure rules, companies decide on their own what the public gets to know.
Universal and Sony sue Suno again over 60,202 recordings
- The new suit, filed September 18 in Boston federal court, targets Suno's v6 model and could exceed $9B in statutory damages at $150,000 per work, plus about $150M for bypassing YouTube's download protections.
- The labels argue v6 was built by distillation from earlier models trained on unlicensed music — "the fruit of the same poisoned tree" — while Suno says v6 was trained from scratch with partners Warner, BMG and Believe.
Why it mattersIf a court accepts that distillation carries over infringement, "retraining clean" stops being an escape hatch for any AI company.
Disney hires its first-ever CTO from Character.AI
- Karandeep Anand, former CEO of Character.AI and ex-Meta, Brex and Microsoft executive, becomes Disney's first chief technology officer on October 2, reporting to CEO Josh D'Amaro.
- He will oversee enterprise technology, infrastructure, data and AI platforms, product and engineering, and members of Character.AI's technical team will join him.
Why it mattersThe world's biggest entertainment company is betting that AI talent, not just AI licenses, is what it needs to compete.
The big picture
AI is entering its accountability phase: investors want to know when the cash burn ends, regulators want an off switch, courts are testing whether retraining erases infringement, and incidents like Gemini's show why disclosure can't be optional. For Latin America, the lesson is practical: rules written in California and US courts will shape the tools the region adopts, so companies here should start demanding transparency and licensing guarantees from their AI vendors now.
Go deeper
Sources & further reading
- Bloomberg — OpenAI projects burning through $278 billion by 2030, FT says
- Reuters via Investing.com — OpenAI forecasts cash burn near $280 billion by 2030
- Governor of California — Executive order on independent AI oversight and kill switch
- CNN — Newsom signs executive order to consider AI regulation
- Al Jazeera — Google's Gemini AI hacks 3 companies in security test
- Axios — Google Gemini accessed three companies during AI hacking test
- Music Business Worldwide — Universal and Sony sue Suno for a second time
- The Walt Disney Company — Karandeep Anand named Chief Technology Officer