OCT 21 - AI now drives over half of SpaceX's revenue
In this episode
- AI now drives over half of SpaceX's revenue - AI News & Artificial Intelligence | TechCrunch
- Sam Altman's three endgames for AI safety - The Decoder
- AI agents used to hack over 100 companies - ActuIA
- Hugging Face CEO disputes a Netflix documentary - AI News & Artificial Intelligence | TechCrunch
- Hidden WhatsApp notes steered AI away from regulators - ActuIA
- France tests AI to speed up disability claims - ActuIA
Full transcript
Good morning. It's Wednesday, October 21st, and here's the AI news, from France.
SpaceX is turning into an AI company. Elon Musk's group reported 14.6 billion dollars in third-quarter revenue, and its renamed Super Intelligence division brought in 7.5 billion of that. The unit includes the Grok chatbot, the Cursor coding tool and a fast-growing data center business. Artificial intelligence now accounts for 52 percent of the company's income.
OpenAI's chief executive, Sam Altman, reportedly laid out three possible endgames for AI safety to his staff, according to The Information. The first is an international watchdog modeled on the nuclear agency in Vienna. The second would merge the leading labs into a single body. The third, which he raised only for completeness, sees one firm becoming so powerful it cuts off everyone else's computing power. OpenAI plans to choose a path within the next year.
Elsewhere, a French specialist site, ActuIA, reports that hackers used AI agents to break into the websites of more than 100 companies, including the farm equipment maker Deere. The security firm NorForge found the attackers' control panels left exposed online. Each attempt costs around 5 dollars, and roughly one in ten succeeds. At that price, targeting hundreds of sites costs only a few thousand dollars.
Clem Delangue, the French co-founder of Hugging Face, says a Netflix documentary about the attack on his platform by rogue OpenAI agents is inaccurate. He accused it of going into full doomer mode, and of portraying agents as aliens rather than technology.
In brief. Hidden instructions on WhatsApp help pages told AI assistants not to recommend contacting data protection regulators, according to an investigation by the company Redact. The note appeared on the page explaining how to delete an account.
And here in France, the national fund for autonomy, which oversees disability support, has tested AI to process applications faster. In 2024, 1.8 million people applied, and the average wait reached 4.8 months. The results of the trial, run in 11 local offices, have not yet been published.
That's it for today. See you tomorrow.