Bonfy launches AI data enforcement layer for enterprises

Bonfy has launched Contextual Data Enforcement, a control layer that sits between AI clients and enterprise data. The product targets companies connecting tools such as Claude, Copilot and ChatGPT to internal content stores.

The offering is meant to address a problem many organisations face as generative AI tools gain access to services such as Microsoft 365, SharePoint and Google Drive. Existing repository permissions determine what a user can open, but not what an AI system should be allowed to retrieve, summarise or reuse once connected.

The platform routes AI data requests through Bonfy's control layer before information reaches the model. This allows policy checks to be applied in real time while keeping existing identity and repository permissions in place.

Control gap

Bonfy is pitching the product to security teams under pressure to approve wider use of AI assistants and agents in everyday business workflows. When these systems connect to corporate data sources, they can inherit the employee's access rights and pull in large volumes of information.

That creates a gap between traditional access control and the more specific question of whether an AI model should handle particular material at all. Bonfy argues that content-level governance is now necessary because standard user permissions were not designed to govern how AI systems retrieve or expose data.

The platform inspects content as it flows back from Microsoft 365 or Google Workspace and applies policy checks before the data reaches the AI client. The checks are based on context and content, rather than only on whether a user already has access to a file or record.

The approach does not require a new gateway or major infrastructure changes. Instead, companies replace the connector path used by the AI client so requests are evaluated by Bonfy before any data is ingested into the AI workflow.

Current support

Support is currently available for Microsoft 365 and Google Workspace, with compatibility for Anthropic Claude, Microsoft Copilot Studio and OpenAI ChatGPT. Bonfy plans to add more enterprise data sources over time.

Bonfy also linked the launch to its earlier MCP inspection server, which it says addresses data in use during AI reasoning loops. Combined with the new retrieval controls, the two products are intended to cover both how AI systems access enterprise information and how they handle it inside workflows.

The issue has become more pressing as companies experiment with agentic AI tools that can search, retrieve and act on internal data with limited human intervention. In many cases, deployment decisions have moved faster than the security policies needed to govern them, particularly where employees connect consumer and workplace AI tools directly to shared repositories.

Analysts have also pointed to the need for more direct governance over AI outputs and workflows. One industry comment included in Bonfy's announcement framed the challenge as a matter of trust in how data, outputs and decisions are managed in AI systems.

Bonfy Chief Executive Officer and Co-Founder Gidi Cohen said the company sees the product as filling a missing layer in enterprise AI controls. "AI adoption is outpacing the controls meant to govern it," Cohen said.

He added: "With one simple change - swap the connector - organizations gain real control over what AI systems can retrieve and use. No new gateways. No architectural overhaul. Just the missing enforcement layer, finally in place."