Post by Y Combinator

Silmaril (YC P26) is the first self-healing prompt injection defense. It catches 2x more attacks at a 10x lower latency than leading guardrails. AI-enabled hackers are generating new types of attacks faster than any security team can keep up with. Current defenses rely on recognizing known patterns. They catch, at best, 61% of real-world attacks. That's almost a coin toss. Security teams are stuck manually updating rules while attackers have already moved on. Silmaril understands your application’s attack surface and retrains continuously, blocking novel attack patterns in under an hour. It protects your entire AI stack: inputs, tool calls, MCP, connectors, and internal agents like Claude Code and OpenClaw. In the last 2 months, Silmaril has prevented 8 breaches, blocking $28M in damages. Congrats on the launch, Aum Upadhyay and Eduardo Velasco! https://www.silmaril.dev/ https://lnkd.in/ga9vBqPK

Video Content