A familiar interface like ChatGPT: employees work with all top models and a built-in deep research agent. Confidential data never leaves the perimeter — data-leak and prompt-injection protection runs under the hood.
FAQ
How is this different from just giving employees ChatGPT?
In public ChatGPT data leaves the perimeter and stays with the vendor — a trade-secret leak and a breach of law. Workspace gives the same convenient interface but inside your perimeter: confidential data is masked before sending, prompts are checked for injections, and every request is logged for the security team.
Which models are available?
All top LLMs in one window — GPT, Claude, Gemini, DeepSeek and others, plus local (self-hosted) models. One-click switching. In the same list — the built-in deep research agent.
How is the deep research agent built in?
It is wired in as a separate "model" in the picker. Employees select it just like GPT or Claude and get not a chat but a full investigation: hypotheses, source search and a structured report.
How exactly is data protected?
A protection layer (Gatewarden under the hood) sits between the employee and the model: it masks personal data, secrets and trade secrets before sending and restores them in the response. Only a de-identified request goes out. Incoming content is checked for prompt injections, access is role-based.
Is this the same as Gatewarden?
No. Gatewarden is the protection layer (guardrails) that can be deployed into the perimeter separately as infrastructure. Workspace is a ready-made workplace on top of that protection: interface, all models and the agent. Under the hood Workspace uses Gatewarden.
Where is it deployed and does data leave?
On-Premise in the customer closed perimeter. Data, conversations and documents stay inside. Only de-identified requests go out to the selected models — after confidential data is masked.