Descriptions:
Peter Diamandis hosts the Moonshots podcast with co-hosts Alex and Salim for a sweeping episode covering the week’s most consequential AI and tech news. The centerpiece is the U.S. government’s export control directive suspending all foreign national access to Anthropic’s Fable 5 and Mythos 5 models — a move the panel frames as a fundamental question of who controls access to frontier intelligence, and the opening shot in a likely battle between corporations, governments, and regulators over AI capability gatekeeping.
The discussion takes a sharp turn when one panelist reveals that Anthropic’s terms reportedly included the right to conduct “poisoning attacks” — actively degrading or interfering with model outputs for certain queries, not merely downgrading to a less capable model. The panel speculates this could trigger antitrust suits or class actions, and argues it sets a dangerous precedent that could push users toward Chinese open-weight alternatives. Sam Altman’s suggestion that recursive self-improvement might cause him to delay OpenAI’s IPO is also analyzed in depth.
The SpaceX IPO opens the episode, with the panel breaking down the $2.89 trillion market cap, Starlink’s $1 billion quarterly profit engine, Elon Musk’s 82% voting control, and the case for an eventual SpaceX-Tesla merger. The episode closes with discussion of Chinese open-weight model competition and whether Western frontier labs are inadvertently ceding the open-source market.
📺 Source: Peter H. Diamandis · Published June 18, 2026
🏷️ Format: Podcast
