AI Insiders spotlights AI security and virtual cell debate at SF Tech Week

AI Insiders brought nearly 200 founders, investors, security leaders and researchers to San Francisco during SF Tech Week 2026 for two nights focused on AI agent security and virtual cell AI. The gatherings highlighted a key question for both fields: whether AI outputs can be trusted as models move from prediction to action.

Why it matters: - AI agent security is moving from a niche research topic to an operational risk for companies deploying autonomous systems. - Virtual cell AI is emerging as a major bet in AI biology, but the event framed prediction as only part of the problem. - Both discussions centered on a practical question: what can be trusted when AI systems start making decisions that affect software, security, drugs or health outcomes?

What happened: - AI Insiders hosted two SF Tech Week 2026 events in San Francisco: Securing the New AI Supply Chain at Pebblebed on October 6 and AI × Longevity at Citizens Private Bank on October 8. - The two nights drew nearly 200 founders, investors, security leaders and researchers. - More than 60 security leaders, founders and investors attended the Pebblebed event. - More than 130 founders, investors and researchers from over 100 companies attended the Citizens Private Bank event.

The details: - Merve Isler, founder and CEO of Marvelous, opened Securing the New AI Supply Chain with a fireside chat with Matt Wise, CEO and co-founder of Archipelo. - The session focused on what happens after an agent acts, including the need for verifiable proof of what an AI agent executed and changed. - Kashish Mittal, AI Security Lead at Salesforce, moderated a panel on securing the AI supply chain. - Jack Cable, CEO and co-founder of Corridor and a co-author of the Transluce research, said secure by design cannot remain a slogan. - Daniel Begimher, CEO and co-founder of Armorer Labs, previously led AWS's first customer-facing AI security agent. - Harsh Verma, Principal Software Engineer for AI and Cybersecurity at Palo Alto Networks, focused on agent and identity security inside large enterprises. - The Pebblebed audience included security engineers and AI researchers from Apple, Microsoft, Nvidia, Okta and Adobe, plus investors from NTT Docomo Ventures and CyberAgent Capital. - Biohub expanded its Virtual Biology Initiative into a $1.8 billion effort the day before AI × Longevity, counting new and existing funding from the US Department of Energy, the National Institutes of Health, Google DeepMind, Meta and Isomorphic Labs. - Ronjon Nag of R42 Group and Stanford Medicine delivered the keynote at AI × Longevity. - Hannah Tono of the Computational Design Institute moderated a panel with Christin Glorioso, CEO of NeuroAge Therapeutics; Jacob Peters, co-founder and chairman of Superpower; and Girish Redekar, CEO of Sprinto. - The panel argued that a model predicting biology well has not proven anything yet, and every result needs a clear decision it changes for a named person. - The panel also said most of the dozens of biomarkers people now receive from tests are not actionable. - The panel said the bottleneck has shifted from AI models to experimental data.

Between the lines: - The security event reflected a broader shift in AI from chat interfaces to systems that can take actions, which raises the stakes for identity, logging and verification. - The longevity event suggested the field is moving from model performance toward evidence, experimentation and clinical utility. - The turnout from operators, researchers and investors shows both categories are attracting interest beyond technical circles.

What's next: - Since the event, AI Insiders has made introductions for AWEAR and YOU(th), the two top-voted startups, to people they wanted to meet from the room. - AI Insiders is continuing to use its event network to connect founders, investors and operators across cybersecurity, fintech and payments, robotics and automation, healthcare and biotech, media and content, and government and policy. - AI Insiders is an invite-only AI ecosystem that started in San Francisco and can be applied to at AI Insiders.

The bottom line: - The two nights showed how AI security and AI biology are converging on the same demand: proof, not just prediction.