News of the day
1. OpenAI allegedly uses a custom ChatGPT to scan internal communications like Slack and emails, aiming to identify employees leaking information. This raises privacy and trust concerns. → Read more
2. Anthropic will cover consumer electricity costs for new data centers, invest in power generation, and cap peak energy consumption. → Read more
3. Uni Systems Luxembourg leads the HORIZON cPAID project, developing a framework to make AI secure by design. It addresses new cyber threats across critical sectors. → Read more
4. Chinese AI companies are leading in open-source model development, offering high-performance, low-cost alternatives that are rapidly gaining global traction and reshaping AI innovation. → Read more
Our take
Hi Dotikers!
The AI That Watches Those Who Build It
This week, a revelation from The Information reminds us that irony is alive and well. OpenAI is using a special version of ChatGPT to track down employees suspected of leaking information to the press. The setup is formidable: the security team feeds published articles into a custom-built ChatGPT connected to Slack messages, emails, and internal documents, capable of cross-referencing in minutes what used to take weeks of investigation. The system draws on an internal architecture spanning over 600 petabytes of data across 70,000 datasets.
It is hard not to see a symbol here. Here is a company founded in 2015 as an open, nonprofit research lab, one that trained its models by scraping the open internet, now deploying its own artificial intelligence to surveil the very people who build it. The official mission remains "AGI for the benefit of all humanity." The method, however, looks more like a digital panopticon.
And this is not an isolated case. In 2024, exit agreements threatened former employees with losing millions in equity if they dared criticize the company. Several of the most respected AI safety researchers left OpenAI denouncing a culture where "safety has taken a backseat to shiny products," in Jan Leike's words. Daniel Kokotajlo gave up nearly two million dollars to regain his freedom of speech. More recently, a vice president in charge of product policy was fired after opposing the launch of an "adult mode" in ChatGPT.
The real question may not be who is leaking, but why so many people feel the need to.
M.
Meme of the day
