Hasabis' "Citadel Plan": New Book Reveals How DeepMind's Security Vision Crumbled

According to 1M AI News monitoring, The Atlantic has published an excerpt from the new book The Infinity Machine by technology historian and financial historian Sebastian Mallaby, based on continuous interviews from 2023 to 2026, reconstructing the evolution of Google DeepMind co-founder Demis Hassabis's AI security philosophy since Google's acquisition in 2014.

In his early years, Hassabis envisioned building superintelligence safely for all humanity through a single scientific team, even planning to retreat top researchers to a secret stronghold at a critical moment, similar to the Manhattan Project. A researcher who joined the team recalled that towards the end of the interview, Hassabis warned him to be mentally prepared to "fly to a secret location in Morocco at any time."

When Google acquired DeepMind in 2014, Hassabis set out rare conditions: establishing an independent external oversight committee, prohibiting military applications, and preventing Google from full control of technology deployment. Google accepted. The following year, he convened a secret meeting at Musk's headquarters in Hawthorne, California, in an attempt to bring potential competitors into a united front, but it backfired: Musk subsequently partnered with Sam Altman to establish OpenAI.

Subsequently, Hassabis launched a secret operation codenamed "Project Mario," trying to regain governance autonomy from Google, forming a legal team, securing a $1 billion funding commitment from LinkedIn co-founder Reid Hoffman, and considering spinning off DeepMind from Google. This three-year battle ultimately failed. Co-founder Mustafa Suleyman was also forced to leave the company in 2019.

At the end of 2022, after the release of ChatGPT, Hassabis completely abandoned the high-minded path of only doing beneficial science. He told Mallaby, "This is wartime," and then fully engaged in the competition between Gemini and ChatGPT. Hundreds of billions of dollars poured into the AI arms race, where neither national regulation nor corporate governance structures could stop the competition. Hassabis's security philosophy also underwent a fundamental transformation: "Security is not about governance structure. Even if there is a governance committee, it is highly unlikely to make the right decision in a critical moment." His new strategy is to ensure that he "has a seat at the decision-making table and can participate in deciding on solutions when security issues arise." At the same time, Google has been actively pitching AI to the U.S. defense system, marking a stark reversal from the military ban Hassabis set for the acquisition.