Develops AI models/platforms like ChatGPT for enterprise, health/science adoption; $20B ARR 2025, focus on practical AI agents/workflows ConstellationR, OpenAI Blog
Develops AI models/platforms like ChatGPT for enterprise, health/science adoption; $20B ARR 2025, focus on practical AI agents/workflows ConstellationR, OpenAI Blog
Responses/Agents API with tools (functions, file/code search, web/computer use); reasoning models for decisions, workflows OpenAI Agents, Tools
Mixpanel leak Nov 2025 (names/emails); Whisper med hallucinations; fake legal cases (2023); resume bias tests OpenAI, Evidently AI, Bloomberg
Detailed breakdown of every risk category for enterprises deploying OpenAI models in agentic AI workflows.
Models fabricate info (Whisper invents med terms/race; ChatGPT fake cases); agents amplify via loops/reliance Evidently AI
GDPR probes on training data; third-party leaks (Mixpanel); agent prompt injection leaks prompts/data AI CERTs, OpenAI
GPT ranks resumes with racial/gender bias (Black names low for finance/SE; fails adverse impact); job-dependent Bloomberg
Prompt injection/jailbreaks bypass guardrails (2024 paper, Guardrails vuln 2025) OpenAI Forum, Gopher
Agents risk data leakage, excessive actions, credential misuse via injection/tool poisoning Zenity, LinkedIn
Context drift/confirmation bias ignores corrections, hallucinates searches OpenAI Forum
EU GDPR probes (Italy training data, breach notice); potential high-risk under AI Act AI CERTs
Lawsuits over training data; compelled 20M ChatGPT logs (2026) NatLawReview
Models aid fraud/phishing; malicious use reports OpenAI Report
Tech E&O for errors/hallucinations; D&O for governance; check AI exclusions in legacy CGL/D&O; emerging AI policies Liberty, Hunton
SoftBank, Target, Lowe’s, ServiceNow TechCrunch