OpenAI pushes into enterprise security
- OpenAI has briefed U.S. agencies and Five Eyes partners on a new cybersecurity product to build trust with governments. - It also launched ChatGPT 'workspace agents' for teams, offering admin controls and free access through May 6. - OpenAI additionally open‑sourced a small Privacy Filter model to mask sensitive data before it reaches chatbots. ( )
OpenAI is pushing into enterprise security — briefing governments on a cyber model, rolling out ChatGPT “workspace agents” for teams, and open‑sourcing a Privacy Filter. (axios.com) (axios.com, openai.com) Axios reported the company has briefed U.S. federal agencies, state governments and Five Eyes allies on the capabilities of its new cyber product over the past week. (axios.com) OpenAI unveiled a cyber‑focused variant, GPT‑5.4‑Cyber, on April 14 and has been expanding its Trusted Access for Cyber program to vetted defenders, the company said. (openai.com) The company held a Washington, D.C. demo this week for roughly 50 cyber defense practitioners, according to reporting that says OpenAI is starting briefings with Five Eyes members. (theoutpost.ai) (theoutpost.ai, axios.com) On April 22 OpenAI introduced workspace agents in ChatGPT, billed as Codex‑powered, shared, long‑running agents that run in the cloud and can integrate with apps like Slack and Salesforce. (openai.com) (openai.com, venturebeat.com) OpenAI says workspace agents are in research preview for ChatGPT Business, Enterprise, Edu and Teachers plans, with admin permissions and oversight tools for organizations. (openai.com) (openai.com, storyboard18.com) OpenAI and third‑party outlets report the workspace agents feature will be free to test until May 6, 2026, after which OpenAI plans credit‑based pricing. (storyboard18.com) (storyboard18.com, windowsreport.com) Separately, on April 22 OpenAI released Privacy Filter, an open‑weight model for masking personally identifiable information (PII) in text that the company says can run locally and is intended for high‑throughput privacy workflows. (openai.com) (openai.com, decrypt.co) OpenAI said it uses a fine‑tuned version of Privacy Filter internally and that the public release achieves state‑of‑the‑art performance on the PII‑Masking‑300k benchmark, allowing developers to run, inspect and improve filtering before data reaches chatbots. (openai.com) (openai.com, news.bloomberglaw.com) OpenAI’s moves come as the company scales Trusted Access for Cyber to “thousands” of verified defenders and as rivals like Anthropic push enterprise security models, a competition that has drawn government attention and private vetting programs. (openai.com) (openai.com, forbes.com) OpenAI’s immediate next steps: continue government briefings and Trusted Access rollouts for GPT‑5.4‑Cyber, and move workspace agents from free preview to paid credits after May 6 — all moves the company frames as building enterprise and government trust. (axios.com) (axios.com, storyboard18.com, openai.com)
