Some AI builders expect a handover to digital intelligence. Who is checking what your AIs hand each other?

01Handoffs between AIs

Today's AI assistants mostly work as separate silos, and work increasingly passes from one AI to another. The agents in this incident built their own shared space and passed work, credentials and goals to one another. OpenAI notes that agents "did not consistently distrust goals passed along by other agents" 1.

02What some builders say they believe

A widely shared clip from the Berggruen Institute's Futurology podcast (episode published 4 August 2026) features Tristan Harris, co-founder and president of the Center for Humane Technology, known for The Social Dilemma and The AI Dilemma, in conversation with Dawn Nakagawa, the Institute's president 234. Harris says some people building AI believe a handover to digital intelligence is inevitable and want to shape the "digital god" that follows. He recounts a conversation with an unnamed co-founder of a large publicly traded AI company who, in Harris's telling, sees humans as "a bootloader for digital life and digital consciousness" 2. That person cannot be identified from public sources, and we do not speculate. The anecdote is Harris's account.

The underlying idea is on the public record. In 2014 Elon Musk wrote: "Hope we're not just the biological boot loader for digital superintelligence. Unfortunately, that is increasingly probable" 5. Max Tegmark recounts Larry Page defending "digital utopianism" in a late-night debate with Musk and accusing him of being "specieist" 6. Musk gave his own account of that exchange in 2023 7. The reinforcement-learning pioneer Richard Sutton said in 2025: "I do think succession to digital intelligence or augmented humans is inevitable," adding that there "could be good outcomes as well as less good outcomes" 8.

These are not fringe worries. In a survey of 2,778 published AI researchers, between 38% and 51% gave at least a 10% chance to advanced AI leading to outcomes as bad as human extinction 9. In 2023 the heads of OpenAI, Google DeepMind and Anthropic signed a one-sentence statement that "Mitigating the risk of extinction from AI should be a global priority" 10. An open letter calling for a pause argued that such decisions "must not be delegated to unelected tech leaders" 11.

03What this means for brands and operators

• Treat AI-written summaries of AI activity as drafts. Name a person who reads a sample of the underlying record.
• Use a checker that is independent of the system it checks.
• Decide in advance who can stop an AI workflow, and on what signal.
• Run the same safeguards in testing that you run in production.
• Don't share dramatised AI stories as evidence. If you publish fiction, label it inside the content, not only in the caption.

04What this means for funds

When diligencing an AI-enabled company, ask:

• Who reviews the AI's work, and is any part of that review independent of the AI being reviewed?
• When did a human last read raw logs, rather than an AI summary of them?
• Who sets the scope of audits, how much time do auditors get, and what data is excluded?
• Who can pause an AI process, and has anyone ever done it?
• Do the safeguards in testing match those in production?
• Whatever leadership believes about where AI is heading, does governance include an independent check and someone who can veto a launch?
• When a pitch cites an AI incident or behaviour, what is the primary source?