Artificial intelligence news.
Straight from the original announcements.

Google OAuth adds DPoP support to protect refresh tokens from theft

October 15, 2026By the AI News Bites Editorial TeamSource: Google for Developers
Google OAuth adds DPoP support to protect refresh tokens from theft
Image: Google for Developers

Google has added support for DPoP (Demonstrating Proof-of-Possession) to its OAuth Platform, binding refresh tokens to a client-generated keypair so that a stolen token cannot be replayed without the private key. According to Google, access tokens remain standard Bearer tokens for performance reasons.

Key points
  • Google's OAuth Platform now supports DPoP (RFC 9449) binding for refresh tokens.
  • Access tokens, typically valid for 1 hour or less, remain standard Bearer tokens to avoid per-request verification overhead.
  • DPoP does not protect session cookies; Google and WICG are developing Device Bound Session Credentials (DBSC) for that.
  • For multi-agent delegation, Google suggests Google Cloud Secure Token Service because oauth2.googleapis.com lacks RFC 8693 support.
  • Google published a DPoP Adoption Guide and a SKILL.md to help implementation.

What DPoP changes for OAuth tokens

Google's developer blog says traditional OAuth bearer tokens work on a "finders-keepers" model. That leaves them exposed to Infostealer malware such as RedLine and NodeStealer, and to Adversary-in-the-Middle (AiTM) attacks.

DPoP is an application-layer OAuth 2.0 security profile defined in RFC 9449. It cryptographically binds tokens to an asymmetric keypair generated by the client. Even if a token is stolen, it cannot be replayed without the corresponding private key, according to Google.

Google also published a DPoP Adoption Guide and a DPoP adoption SKILL.md to help developers implement the profile. The company notes that some optional features and implementation decisions in RFC 9449 are delegated to Google as an authorization provider.

Refresh tokens bound, access tokens unchanged

Google's OAuth Platform supports DPoP binding specifically for refresh tokens. The company calls this a pragmatic engineering tradeoff: sender-constraining the long-lived refresh token neutralizes the primary vector for persistent offline impersonation.

Short-lived access tokens, typically valid for 1 hour or less, stay standard Bearer tokens. Google explains that verifying asymmetric signatures such as ES256 takes significant CPU cycles. If every call to high-throughput resource servers like Gmail, Calendar or Drive required a DPoP proof check, the compute overhead at the API Gateway layer would be massive, with measurable latency and scaling costs.

The result, in Google's words, is that most API requests incur zero cryptographic verification overhead, while an exfiltrated refresh token is useless to an attacker.

Developers should expect extra work. DPoP requires client-side key generation, strict timestamp and nonce handling, computation of specific claims (htm, htu, iat, jti), exact header construction and DPoP proof JWT construction. Google says these changes may add system constraints for secure computation, handling and storage.

Architecture guidance and limits

Google points to the IETF's Best Current Practice for OAuth 2.0 Security (RFC 9700), which discusses sender-constrained refresh tokens. It also cites the IETF draft on OAuth 2.0 for Browser-Based Applications, which describes three models: Backend-for-Frontend (BFF), Token-Mediating Backend and Browser-based OAuth Client. Google recommends using DPoP to sender-constrain refresh tokens in all of them.

The company contrasts DPoP with transport-layer binding such as mTLS (RFC 8705), which it says often breaks or terminates at modern cloud proxies and load balancers. Because DPoP works at the HTTP header level, Google says it is resilient in multi-hop environments, including emerging Model Context Protocol (MCP) and autonomous agent architectures.

DPoP has boundaries. It does not protect standard web application session cookies, which remain a primary target for Infostealer malware. Google and the Web Incubator Community Group (WICG) are developing Device Bound Session Credentials (DBSC), which binds browser session cookies to device hardware such as a TPM. Google describes the two together as a defense-in-depth strategy covering both API and web sessions.

Multi-agent workflows and what it means for developers

Multi-agent chains pose a specific problem. If Agent A delegates a task to Agent B, which calls Resource C, Agent B cannot sign a DPoP proof for Agent A without Agent A's private key, and sharing private keys violates the core security principles of DPoP.

Google's consumer OAuth endpoint (oauth2.googleapis.com) does not support OAuth 2.0 Token Exchange (RFC 8693). For multi-agent or federated cloud workloads, the company suggests Google Cloud Secure Token Service (STS). Through STS, an intermediate agent can swap the upstream token for a new, short-lived federated token bound to its own local key pair, keeping the sender-constrained model intact without exposing keys.

Google also says Workspace Context-Aware Access (CAA) uses DPoP to secure enterprise AI workloads on the Gemini Enterprise Agent Platform, validating agent identities at the Agent Gateway.

For developers integrating Google APIs, the message is that long-lived refresh tokens can now be protected against theft, at the cost of more client-side cryptographic work. Teams handling sensitive credentials, or building agent and MCP systems, are the audience Google targets with the guide.

GoogleGoogle OAuthDPoPGemini Enterprise Agent PlatformGoogle Workspace Context-Aware Accesscybersecurityenterprise AIAI agents

Related news

Google reports Agent Anomaly Detection benchmark results
Safety and ethicsNovember 12, 2026

Google reports Agent Anomaly Detection benchmark results

Google says its Agent Anomaly Detection feature flagged 87% of harmful instructions that AI agents carried out in tests against five public, non-Google-written security…

Google Gemini 4 Carbon Comes to n8n With a Focus on Safer Agents
Products and toolsNovember 10, 2026, 17:43

Google Gemini 4 Carbon Comes to n8n With a Focus on Safer Agents

Google Gemini 4 Carbon is now available in n8n, giving users access to a model that Google describes as its strongest yet. n8n says the model is suited to automation…

NVIDIA Gives OpenShell and AI Infrastructure Tools to Open Source
For developersNovember 10, 2026, 17:00

NVIDIA Gives OpenShell and AI Infrastructure Tools to Open Source

NVIDIA is contributing its OpenShell runtime for securing autonomous agents to the Cloud Native Computing Foundation, alongside work on Agent Substrate and new AI…