AI Hacks Back: Build Fast and Safe

When AI Gets Hacked, Founders Pay First. AI founders are shipping at model speed, and the newest attacks move even faster, slipping through prompts, agents, plugins, and “harmless” automations that look like normal product behavior until it’s too late. In this high-signal panel, we’ll break down what’s already hitting startups in 2025, including the Salesforce prompt-injection breach, the Anthropic agentic attack, antigravity-style risk scenarios, and the hidden failure modes of vibe coding, and the new wave of AI-native intrusions that don’t look like the threats we grew up with.