Sriram Krishnan joins Theo Jaffee and Sofia Puccini to discuss a wave of open-weight model releases. He argues that more model choices can put pressure on frontier pricing and shift value toward serving infrastructure and application harnesses. These are his economic expectations, not guarantees that a particular model matches the frontier or that prices will fall.
Krishnan separates support for open models from concerns about national origin, defensive access and unauthorized extraction of model outputs. He argues that inspectable weights can benefit security and that legitimate developers need clearer conditions for distillation. The interview does not establish that openness makes every model secure, and its policy discussion is not legal advice.
On automated AI research, Krishnan favors competition and responses to concrete risks over a single predicted acceleration curve. He describes using AI to strengthen defenses and argues that useful model products can support the wider hardware and energy stack. He explicitly speaks as a former adviser without inside information about current government decisions.
Watch on YouTube
