Uploaded April 2026 | Updated September 2026, 13 hours ago

---

"English is the new way to hack." In this episode from the WorkOS booth at HumanX, Michael Grinich sits down with Noam Schwartz, CEO and Co-Founder of Alice (formerly ActiveFence), to discuss the terrifying and fascinating world of AI Trust & Safety.

Noam shares his experience red-teaming the world’s most advanced foundation models and explains why AI security is moving closer to "fraud detection" than traditional cybersecurity. They explore the rise of prompt injection, the vulnerability of open-source models without guardrails, and why the industry is rapidly shifting toward a "multi-model" architecture to mitigate risks.

Key Highlights:
• Why "Prompt Injection" is the AI era's version of SQL injection.
• The "Superposition" of AI: How to prevent models from becoming weapons or targets.
• Multi-modality risks: Beyond text—securing real-time video, audio, and deepfakes.
• Prediction: Why hyperscalers are about to consolidate the foundation model industry.

Learn more about WorkOS: workos.com Secure your AI with Alice: alice.ai