Snapshot 253082
Normalized text
Scripts and page chrome removed; this is what change detection compares.
TrustSecurity Identity, credentials and policy checks for every AI agent Guild registers each agent with an owner, keeps your keys out of agent code, checks every tool and model call against your policy, and logs every action. Your security team sets the rules. Guild enforces them at runtime. Book demoVisit the trust center SOC 2 Type 2 audited Unqualified audit opinion 77 controls tracked 8 subprocessors listed Annual penetration test AI agents hold credentials and call production systems. They act at machine speed, often with more access than their job needs. Most teams can’t say what each agent can reach, or prove what it did. Standing access API keys pasted into code and prompts give an agent more reach than its task needs. Unregistered agents Agents built outside review run with no owner, no registry and no limits. No audit trail When something goes wrong, there’s no record of which agent did what, or why. How Guild secures agents at runtime Every agent runs inside Guild’s governed runtime, which checks each action before it happens rather than scanning agents after the fact. Security architecture Every agent is registered with an ID and an ownerEach agent gets a unique ID and is tied to the person who created it, so you always know what’s running, where, and who answers for it. Credentials are added server-side, never given to the agentGuild adds credentials after a policy check. They never enter the agent’s code, container or prompt, and each agent gets only the operations its job needs.How credentials work Every tool and model call is checked against policyAgents run in network-isolated containers with no general internet access. Every outbound call goes through Guild and is checked against your policy. Deny wins over allow.Credential policies Risky actions wait for a person to approve themRoute sensitive actions to a human before they run. Stop a session or revoke a credential at any time, and it takes effect on calls already in flight. Every agent action is written to an audit logA tamper-evident audit log records every admin change. A session log records every model call, tool call and error. Export both for your next review.Audit logs SOC 2 audited, encrypted and penetration tested Guild’s own security program is independently audited, and every document a security review asks for is in our trust center. Visit the trust center SOC 2 Type 2 Examined by an independent auditor, who issued an unqualified opinion. Request the report Encrypted at rest and in transit Datastores holding customer data are encrypted at rest. Data sent over public networks is encrypted in transit. See the controls Penetration tested every year Penetration testing runs at least once a year, and every finding is fixed against a set deadline. See the controls A published Data Processing Agreement Formal retention and disposal procedures, and customer data is removed from the service when you leave. Read the DPA Every subprocessor listed Guild runs on Google Cloud. Each third party that processes data, model providers included, is listed with what it does. See subprocessors Report a vulnerability Report a vulnerability or ask a security question at support@guild.ai. Email support@guild.ai Get the SOC 2 report, DPA and subprocessor list The SOC 2 Type 2 report, DPA, controls and subprocessors are all in our trust center. The report is shared under NDA. Visit the trust center Approve credentials and guardrails once, and every team reuses them Sign off on credentials, integrations and guardrails once, and every engineer builds on them. Compliance reviews get faster and your review queue gets shorter. Explore Guild for CISOs Guild security FAQ Answers for your review. Visit the trust center Guild gives every agent a unique ID and an owner, adds credentials server-side so agents never hold your keys, runs agents in network-isolated containers, checks every tool and model call against policy, routes risky actions to a human, and records every action in an audit log. Yes. Guild completed a SOC 2 Type 2 examination and received an unqualified opinion from its independent auditor. Request the report through the Guild trust center; access requires an NDA. No. Guild adds credentials server-side, after a policy check, so they never enter an agent’s code, container or prompt. Each agent gets only the operations its task needs. Not directly. Guild runs agents in network-isolated containers with no general internet access. Every outbound tool and model call goes through Guild and is checked against your policy, and deny rules win over allow rules. Guild keeps a tamper-evident audit log of every admin change and a session log of every model call, tool call and error. Both can be exported for audits and security reviews. Guild runs on Google Cloud Platform. Every subprocessor that handles customer data is listed in the Guild trust center with what it does. Anthropic is the default model provider for Guild’s AI features, with Google Gemini and OpenAI as alternatives. All three are listed as subprocessors in the Guild trust center. Yes. Guild encrypts datastores that hold customer data at rest, and encrypts data in transit over public networks. Guild follows formal retention and disposal procedures and removes customer data from the service when you leave. The terms are in the Guild Data Processing Agreement (DPA), published at guild.ai. Yes. Guild runs penetration testing at least once a year, and every finding gets a remediation plan fixed within set deadlines. Email the details to support@guild.ai. Security questions about Guild go to the same address. See how Guild secures AI agents in production Book a demo and we’ll walk through identity, credentials, policy and audit on your own agents. Contact salesVisit the trust center