Prompt Security vs Promptfoo

Prompt Security is an AI security product that controls how employees use AI tools and protects the AI applications and agents companies build. Itamar Golan (CEO) and Lior Drihem (CTO), who had worked at Check Point and Orca Security, founded it in August 2023; it is based in Tel Aviv. It raised $23 million: a $5 million seed round led by Hetz Ventures (January 2024) and an $18 million Series A led by Jump Capital (November 2024), with Ridge Ventures, Okta and F5 joining. SentinelOne, the Mountain View, California security company, agreed to buy Prompt Security on August 5, 2025 and completed the deal on September 5, 2025. Its Form 8-K puts the consideration at approximately $180 million in cash and SentinelOne Class A stock. Prompt Security is still sold under its own name, as part of SentinelOne's Singularity Platform. The product page is on sentinelone.com; prompt.security now hosts research and free tools such as the open-source Prompt Fuzzer, and the old press release we checked there redirects to SentinelOne. **What it does** SentinelOne's Prompt Security page lists three areas. AI Usage Control inventories the AI tools and code assistants employees and developers use, including unsanctioned shadow AI, and redacts sensitive data and enforces policies across more than 15,000 AI services in real time. It also covers GitHub Copilot, Cursor and Claude Code, keeping secrets and intellectual property out of them. AI Application Security tests homegrown AI apps for prompt injection, jailbreaks and data poisoning before deployment, runs a real-time AI firewall, and turns red-team findings into guardrails. Agentic AI Security finds shadow MCP servers and unsanctioned agents, enforces least-privilege access for agents and keeps a searchable audit log. SentinelOne says Prompt's MCP gateway scores the risk of more than 13,000 known MCP servers and can allow, block, filter or redact their actions. Prompt deploys through a lightweight agent and browser extensions, and covers major LLM providers, including OpenAI, Anthropic and Google, and self-hosted models. **Product changes since the acquisition** At its OneCon conference in November 2025 SentinelOne launched Prompt Security for Employees, for AI Code Assistants and for AI Applications, plus Prompt Security for Agentic AI in beta. At RSAC in March 2026 it announced Prompt AI Agent Security, a discovery and governance control plane for agents and MCP servers (agent posture management was in preview), and Prompt AI Red Teaming for testing homegrown AI apps. **Frameworks** Neither SentinelOne's Prompt Security pages nor prompt.security name the EU AI Act, ISO 42001 or the NIST AI RMF; they refer only to meeting compliance requirements in general terms. This profile lists no frameworks. **Who it's for** Companies that want to let employees use generative AI and coding assistants while controlling data exposure, especially existing SentinelOne customers. Named customers include 10x Banking and Carnival Corporation.

Promptfoo makes an open-source command-line tool and library for evaluating and red-teaming LLM applications and agents, and sells Enterprise and On-Premise editions of it. Ian Webster (CEO) and Michael D'Angelo (CTO) founded the company in 2024; it is based in San Francisco. On March 9, 2026, OpenAI and Promptfoo announced that Promptfoo had agreed to be acquired by OpenAI, subject to customary closing conditions. Promptfoo's About page now says "Promptfoo is now part of OpenAI"; neither company has published a closing date. Promptfoo said it would remain open source, keep supporting a range of providers and models, and keep working with existing customers. OpenAI said it plans to build Promptfoo's technology into OpenAI Frontier, its platform for building and operating AI coworkers, once the deal is finalized. The open-source tool runs locally: evals run on the user's machine and talk directly to the model provider. Its docs name OpenAI, Anthropic, Azure, Google and Hugging Face among supported providers, plus open-source and custom models. Promptfoo's site lists six products: Red Teaming, Guardrails, Model Security, MCP Proxy, Code Scanning and Evaluations. Its homepage says red teaming covers 50+ vulnerability types, from injection to jailbreaks, and that the tool is used by 300,000+ developers and 156 of the Fortune 500. Those are Promptfoo's own figures. The docs cover CI/CD use with GitHub Actions, GitLab CI, Jenkins, CircleCI, Azure Pipelines and others. **AI compliance coverage** Promptfoo's red-team docs include framework presets that group its test plugins by standard: the EU AI Act (Article 5 prohibited practices and Annex III high-risk categories), the NIST AI RMF (the MEASURE function, subcategory by subcategory), ISO/IEC 42001 (seven risk domains), the OWASP Top 10 for LLM Applications, the OWASP Top 10 for Agentic Applications, MITRE ATLAS and GDPR data protection testing. Running a preset produces a report that sorts findings by that framework. Promptfoo's own docs say this is testing evidence, not compliance: the EU AI Act page says "EU AI Act compliance requires additional measures" such as documentation, transparency, human oversight and quality management, and the NIST page says automated red teaming should be one part of a wider program. Promptfoo is not an AI inventory, policy or risk-assessment system. Promptfoo's site carries SOC 2 and ISO 27001 certification badges and links to a trust center; it announced SOC 2 Type II and ISO 27001 certification on July 11, 2025. Its pricing page doesn't mention HIPAA or a BAA. **Who it's for** Developers and security teams who want to test prompts, models, RAG pipelines and agents for quality and security problems before release, from the command line or in CI. Larger teams buy the Enterprise or On-Premise edition for shared dashboards, SSO and custom red-team limits.