The risk in malicious AI models: Wiz Research discovers critical vulnerability in AI-as-a-Service provider, Replicate

Wiz Research found a critical tenant-isolation vulnerability on the Replicate platform where a maliciously crafted Cog container enabled root remote code execution and, by injecting packets into a shared network namespace TCP stream to a Redis queue, could intercept and alter other customers' prediction prompts and outputs; the issue was responsibly disclosed and quickly mitigated with no reported customer data compromise.