Contextual Agent Security, a design by Assistant Professor Eugene Bagdasarian with collaborator Lillian Tsai, is now used as a semantic policy engine in Google Cloud's Gemini Enterprise Agent Platform. The approach checks an AI agent's proposed tool calls against the user's intent and plain-language rules before those actions run, addressing prompt injection and intent drift. Bagdasarian also reports that OpenAI confirmed the data-exfiltration attack from his AirGapAgent work in its prompt-injection defenses. Bagdasarian's site