AI Security Summit — San Francisco 2026

The AI Security Summit's San Francisco edition is a single-day gathering pitched squarely at people securing AI systems in production rather than theorising about them. The organisers — a joint initiative from developer-security firm Snyk and the AI Engineer community — frame it bluntly: "one day, two tracks, no filler," with no generic threat overviews or "shift left" platitudes. Held on 15 October 2026 in San Francisco, it sits alongside sibling editions in Paris and London, giving it the feel of a focused circuit rather than a one-off.

The programme splits into a Leadership track and a Practitioner track. The Leadership track targets those who own AI risk at an organisational level — setting budgets, vetting vendors, and writing governance and board-level risk-communication playbooks as the field moves under them. The Practitioner track is for hands-on engineers shipping, testing and breaking AI systems, covering LLM vulnerability patterns, prompt-injection and jailbreak defence, and live AI red-teaming demonstrations. Talks are sourced through an open call, so the agenda skews towards practitioners presenting work they have actually done. If you work on the security of large language models or agentic systems, this is a concentrated day of concrete, attack-and-defence material rather than vendor showcases. Track agendas, speakers and venue details are confirmed closer to the date through the summit's site.