ElevenLabs Agents

Are scope, spend, and authority enforced while the agent runs?

Re-cited 2026-08-25, score unchanged at 2. The original quote ("Define deterministic workflow gates and approval steps.") is absent from the 2026-08-18 capture, and "approval" now occurs zero times on the page - the human-approval half of the claim is withdrawn separately. The guardrail half survives and is what the 2 rested on: the page states "Define guardrails, run simulations, and validate agent behavior with automated tests" and demonstrates a guardrail returning "Request blocked by guardrail" on a financial-advice request. Enforcement is asserted at run time but its scope is not published, so it stays 2.

Judged Aug 5, 2026; evidence newer. We have since re-read Runtime governance, most recently on Aug 18, 2026. No one has re-scored this dimension against it — the score above stands, but it is older than the evidence below it. How current a score is

Is there a tamper-evident record of what it actually did?

Lowered 2026-08-25 from 2. The 2 rested on "All agent actions are logged and auditable so you can validate compliance before and after deployment.", which is absent from the 2026-08-18 capture of the same URL. What the current page supports is conversation-level review - transcripts, conversation history and debug logs - which is a record the customer can inspect, but not a record of all agent actions and with no claim of tamper-evidence. That is a 1, not a 2.

Judged Aug 5, 2026; evidence newer. We have since re-read Audit trail, most recently on Aug 18, 2026. No one has re-scored this dimension against it — the score above stands, but it is older than the evidence below it. How current a score is

Are compliance obligations attached per engagement?

Page lists SOC 2 Type II, ISO 27001, ISO 42001, AIUC-1, HIPAA compliant and PCI DSS Level 1. Certification is asserted but no certificate, auditor name, report date or per-engagement compliance schedule is published, so not a 3.

This score counts Regulatory alignment as undisclosed, and that absence is flagged for review — the cited source may address it. If so, this score is too low. How we check absences

Judged Aug 5, 2026; evidence newer. We have since re-read Compliance certifications, most recently on Aug 18, 2026. No one has re-scored this dimension against it — the score above stands, but it is older than the evidence below it. How current a score is

Does payment depend on a verified result?

No public evidence found of payment linked to a verified outcome. The only pricing terms published on the page are usage-based ("15 minutes included for free") and a commercial licence inclusion. Absence is the finding.

Can the deployment be insured, and is the customer indemnified?

Corrected 2026-08-25 from 0. The prior 0 ("no public evidence found") was honestly recorded against the 2026-08-05 capture, which contains zero occurrences of "insurance". The 2026-08-18 capture of the same URL states: "ElevenAgents is also the first agent platform eligible for AI insurance through AIUC, once certified." That is conditional eligibility through a named certification programme, not cover in force - no carrier, no published limits, no effective date and no customer indemnity - so it clears 0 (which asserts nothing at all is published) but goes no higher than 1. This remains a trust gap and the referral link is unchanged.

Judged Aug 5, 2026; evidence newer. We have since re-read Insurance available, most recently on Aug 18, 2026. No one has re-scored this dimension against it — the score above stands, but it is older than the evidence below it. How current a score is