Unified Governance and Security
Proofpoint and Anthropic protect every Claude conversation on both sides, flagging sensitive conversations that have already happened, and blocking sensitive prompts from ever reaching the model going forward. One consistent line of defense, powered by the DLP policies you already trust.
Most organizations already use Claude, with little visibility into what's already been shared. The Compliance API brings conversations into view, so teams can flag sensitive data and act on it through existing remediation workflows.
The security check runs before inference, so a blocked prompt never reaches the model. It uses the same DLP policies that already secure your email, endpoint, and cloud, so you have nothing to re-author, reconcile, or let drift.
Visibility and prevention both work at the Claude platform level, not the network layer, so protection holds whether a conversation happened on a managed or unmanaged device, with nothing dependent on inspectable network traffic.
the Growing Scale of AI Work
Employees and agents are already sending regulated data, source code, and customer PII into Claude as part of everyday work. Most security teams have no visibility into what's already been shared, and no way to stop sensitive prompts from reaching the model going forward. Closing both gaps means extending the policies you already trust, not standing up a new AI-specific stack. The Compliance API works after the fact, flagging sensitive conversations that have already happened so your team can act on them. Inference Hooks stop sensitive prompts before they reach the model.
Full Visibility
The Claude Compliance API brings existing Claude Enterprise conversations—chats, uploaded files, and projects—into view, so security, IT, and compliance teams can flag sensitive data that's already been sent to Claude and route it into the remediation workflows they already run. It applies the same classifiers, behavioral risk models, and DLP policies already used across email, endpoint, and cloud, plus supervision, retention, and eDiscovery workflows through Proofpoint Digital Communications Governance.
Features
Apply existing classification and behavioral risk models to Claude Enterprise prompts, responses, files, and projects.
Preserve, supervise, and investigate Claude conversations with the same eDiscovery and retention workflows used elsewhere.
Monitor admin events, workspace changes, API key activity, and account settings across Claude Platform.
Inference Hooks
Anthropic's Inference Hooks let Proofpoint check a prompt against your DLP policy before the model processes it. When a prompt carries sensitive data, such as PII, source code, or regulated records, Proofpoint returns a deny verdict, Claude never processes the request, and the employee or agent sees a plain-language message explaining why. If a verdict doesn't return in time, your configured failover decides whether to block the request or let it through unevaluated.
Features
Every prompt is checked against existing DLP policy in the request path before any Claude model runs.
Prompts that violate policy are blocked instantly, with a custom message that turns each block into a teachable moment.
The check runs inside Anthropic's infrastructure, so coverage doesn't depend on a managed device or an inspectable network path.
Both integrations run inside Anthropic's infrastructure, not on your network, so coverage doesn't depend on which Claude surface someone's using, what device they're on, or whether the traffic ever crosses an inspectable network path. Visibility from the Compliance API and prevention from Inference Hooks apply consistently, everywhere Claude does.
Capture Claude Enterprise conversations for review, no matter which surface or device they came from.
Check every prompt before inference, no matter which surface or device it came from.
Request a Demo
Thank you!
A Proofpoint cybersecurity specialist will reach out to you shortly.
Proofpoint integrates with Claude in two ways. The Claude Compliance API brings Claude Enterprise conversations and Claude Platform activity logs into Proofpoint's data security, insider risk, and governance workflows for visibility and investigation after the fact. Anthropic's Inference Hooks let Proofpoint check prompts against the same DLP policy before Claude processes them, blocking sensitive data in real time. Together, they cover both detection and prevention.
Detection, through the Compliance API, gives you the full record of Claude activity so you can review, retain, and investigate after the fact. Prevention, through Inference Hooks, checks a prompt before the model ever processes it and blocks it in real time if it violates policy. They ship together: detection sees everything, prevention stops what matters most before it happens.
The Compliance API integration is available for Proofpoint customers deploying Claude Enterprise and Claude Platform in Anthropic-hosted environments. Inference Hooks currently apply to Claude Enterprise, with availability continuing to expand throughout the year.
Keep them—this integration doesn't replace network-based tools; it covers what they can't see. Enforcement happens inside Anthropic's infrastructure rather than on the network, so it applies the same way whether someone is on a managed laptop or a device you don't control.
Anthropic secures the platform: the infrastructure, model, and hosting. Proofpoint governs your data inside it: Inference Hooks block sensitive prompts before Claude ever processes them, and the Compliance API flags sensitive conversations that have already happened so your team can act on them using the same DLP policies you already run across email, endpoint, and cloud.
