AI Security jobs

AI security covers two overlapping jobs: classic security for companies whose crown jewels are model weights, and the newer discipline of securing AI systems themselves — prompt injection, data poisoning, model extraction, and agents holding credentials they could be tricked into misusing. Plenty of postings quietly want both.

Duties run from hardening training and inference infrastructure to threat-modeling agentic products and building guardrails that survive adversarial users. Security engineers who understand transformers are still rare enough that either half of the pairing gets interviews. OWASP's LLM Top 10 supplies shared vocabulary; most of the rest is being invented on the job.