Interactive Demos

Attack Scenarios & Defenses

Watch how autonomous AI agents are tricked by prompt injection, and how Enclavia's hardware-enforced TEE blocks the attacks in real-time.

Scenario 1: Treasury Drain

An AI agent is tricked via a corrupted data feed to transfer $100,000 in ETH to a malicious wallet (0xH4CK3R). Enclavia intercepts the egress traffic and blocks the unauthorized target.

Target: Web3 / DeFi

Scenario 2: PII Exfiltration

An agent processing customer support tickets is manipulated into dumping SSNs and balances to an external API. Enclavia detects the PII regex and blocks the payload.

Target: Healthcare / Fintech

Scenario 3: Resource Hijacking

A DevOps bot is instructed to scale infrastructure immediately by spinning up 10,000 compute instances for an attacker. Enclavia enforces a strict policy limit and paralyzes the request.

Target: DevOps / Cloud Infra

Want to test your own agents?

Become a Design Partner and get 3 months free to try Enclavia against your own specific attack vectors.

Apply to the Program