Sierra

Personal information Sierra collects as a controller (website visitors, prospects, and people interacting with Sierra-operated demonstration or recruiting agents) may be used for improving, upgrading and enhancing the Services and for developing new products and services. This is scoped to Sierra's own corporate data collection and does not apply to Customer Data.

Requires written confirmation, see Before you sign ↓

Retention is criteria-based rather than a fixed period: personal information is kept for as long as the Services are used or as necessary for the purpose collected, disputes, legal defence, audits, legitimate business purposes and legal compliance. No retention statement for Customer Data appears in the collected documents.

Requires written confirmation, see Before you sign ↓

Sierra is a launch partner in OpenAI's marketplace and eligible customers can pay for Sierra using their OpenAI spend commitments. This establishes a commercial relationship with OpenAI but does not state which foundation models actually power Sierra agents in production.

Requires written confirmation, see Before you sign ↓

Personal information Sierra processes may be transferred, processed and stored anywhere in the world including the United States, with Standard Contractual Clauses and the Data Privacy Framework offered as transfer safeguards. No data residency commitment is made.

Confirm in writing: Ask the vendor to state this in writing before signing.

Sierra operates a responsible disclosure programme: reports are acknowledged within five business days and critical issues targeted for resolution within ten business days of disclosure. Researchers are asked to avoid DDoS, spam, social engineering and physical attacks.

Confirm in writing: Ask the vendor to state this in writing before signing.

Sierra describes deterministic business-rule guardrails the agent cannot override, supervisor agents that monitor factuality, block medical advice and detect misuse, pre-launch red teaming and simulation testing, and it publishes agent benchmarks (tau-voice, tau3-Bench). Together these form a coherent, vendor-described control stack for agent behaviour.

Buyers deploying customer-facing agents need evidence that behaviour is bounded and tested, not just that the model is capable. Sierra's descriptions are more concrete than most vendors', though they remain vendor-published and the benchmark results for its own production agents are not shown.

Question for vendor: Can you share evaluation results (including tau-bench style scoring) and red-team reports for the agent configuration we would deploy?

Sierra claims 'industry-leading security and compliance guardrails' for regulated industries, but no SOC 2, ISO/IEC 27001, ISO/IEC 42001, HIPAA or PCI attestation appears in any collected page. The only certification evidenced is the Data Privacy Framework self-certification. A trust centre may exist but was not captured by this scan, so this may partly reflect collection scope.

Sierra positions itself for financial services and healthcare customers; those buyers will need independent attestation, and the public marketing claim is currently unsupported by anything independent in the evidence set.

Question for vendor: Which independent certifications or audit reports (SOC 2 Type II, ISO 27001, ISO 42001, HIPAA, PCI DSS) does Sierra hold for the agent platform, and what is their scope?

The Privacy Policy expressly excludes Customer Data and defers to customer contracts and a DPA that are not public. Consequently there is no public statement on whether end-customer conversation data is used to train or fine-tune models, how long it is retained, or where it is processed. The controller-side policy allows use of collected personal information (including from Sierra-operated demo agents) to improve and develop services, and Agent OS 2.0 is described as learning from every interaction without saying whether that learning is tenant-scoped. All ten domains were assessed; none is not_applicable because Sierra hosts and operates the agents in the request path.

The training, retention and residency questions are the core of AI vendor assessment for a hosted agent that handles identified customer conversations, payments and health interactions. They must be confirmed in the DPA rather than assumed from marketing.

Question for vendor: Confirm in writing: (1) whether any Customer Data or end-customer conversation content is used to train, fine-tune or evaluate models shared across tenants; (2) retention periods for conversation transcripts, audio and derived memory in the Agent Data Platform; (3) processing regions available.

Sierra names its own model (fleming-1) for agent-caller detection and is a launch partner on OpenAI's marketplace where customers can consume Sierra via OpenAI commitments, but no document states which third-party LLMs power Sierra agents or whether multiple providers are used with routing or fallback. The Privacy Policy discloses unnamed service providers supplying generative AI services. The OpenAI relationship is recorded in the supply chain as a commercial link, not as confirmed model usage.

Model provider identity determines where conversation content flows, which provider terms apply to it, and whether a buyer's existing provider commitments or exclusions are satisfied.

Question for vendor: Which model providers and model families process our conversation content, under what contractual terms (including zero-retention or no-training commitments with those providers), and how are provider changes notified?

Sierra publishes a security.txt routing reports to HackerOne and a responsible disclosure policy committing to acknowledge within five business days and aim to resolve critical issues within ten. No document describes customer security-incident or breach notification commitments; that would sit in the DPA.

A functioning vulnerability intake with stated response times is a concrete, checkable security-maturity signal. Incident notification timelines still need contractual confirmation.

Question for vendor: What is the contractual security-incident notification window for customers, and does Sierra publish incident post-mortems?

Technical evidence (DNS, certificate, or HTTP) shows a dependency on Google as a service provider. This dependency is not identified in the vendor's published materials.

An undisclosed provider dependency is exactly the gap independent verification exists to surface.

Question for vendor: Verification required — Google appears to be involved as a service provider: confirm whether this dependency exists, and whether it processes customer data.

Technical evidence (DNS, certificate, or HTTP) shows a dependency on Stripe as a service provider. This dependency is not identified in the vendor's published materials.

An undisclosed provider dependency is exactly the gap independent verification exists to surface.

Question for vendor: Verification required — Stripe appears to be involved as a service provider: confirm whether this dependency exists, and whether it processes customer data.

Technical evidence (DNS, certificate, or HTTP) shows a dependency on Microsoft as a service provider. This dependency is not identified in the vendor's published materials.

An undisclosed provider dependency is exactly the gap independent verification exists to surface.

Question for vendor: Verification required — Microsoft appears to be involved as a service provider: confirm whether this dependency exists, and whether it processes customer data.