The Cisco AI transparency note set beyond the contact centre

Verified 2026-09-05 · 18 claims · sources tier 2 · 1 disputed

Also known as AI Canvas, AI transparency note, face recognition, gesture recognition, head detection, responsible AI framework.

Field reports suggest that 41 documents carrying the AI Transparency type were observed in a 20 query sample of the Cisco Trust Portal search endpoint taken on 5 September 2026 7. In the field, roughly 2 thirds of these 41 AI transparency documents concern neither Webex nor the contact centre, covering products such as email threat defence, firewall operations, industrial network troubleshooting, and network observability 8.

Publishing structure and portal hosting

Field reports suggest that the observed AI transparency documents are served from 3 different URL paths: a transparency path, a white paper collaboration path, and a privacy data sheet collaboration path 13. In the field, 8 of the 9 notes read in this tranche follow 1 section skeleton running feature overview, model overview, data sources, model evaluation, safety and ethical considerations, fairness, privacy and security, updates and maintenance, and references 1. Every note read carries a 6 principle responsible artificial intelligence preamble and closes with a list of responsible AI principles, the responsible AI framework, and the Cisco Trust Portal 12.

Field reports suggest that 10 of the 41 typed records observed carry a document expiry date already in the past while all 41 are marked active and latest version, and 1 expired note still served a full document 3. Furthermore, 2 of the 9 notes read—the AI Canvas note and the Secure Email Threat Defense note—are marked Cisco Confidential on their pages while the portal classifies them public and serves them from a public path 14.

Cisco is misspelled as Csico in the Vidcast note's document title, its PDF title metadata, its PDF subject metadata, and the portal record, while the body of the note spells it correctly 16. In addition, 3 of the 9 notes carry text copied from a sibling document: the AI Canvas note's PDF subject metadata describes ThousandEyes, the Control Hub note directs administrators to the Vidcast website, and the Calling assistant note's fairness section discusses post call summaries 2.

Models, vendors, and data governance

Model vendors named outside the contact centre are broader than Azure OpenAI and include AWS Bedrock, Amazon Polly, Google Cloud text to speech, Eleven Labs, Deepgram, and named open model checkpoints 15.

None of the 9 notes read publishes a numeric retention period 10. The AI Receptionist note is the only one of the 9 to publish a data residency table, listing 7 regions, and its own footnote qualifies the guarantee for several of them 11.

Restrictions, biometric governance, and fairness

None of the 9 notes read outside the contact centre carries a restriction on using its output for employment decisions 9.

The head detection and face recognition note does restrict use, stating that the feature should not be used for surveillance or for identity recognition beyond the meeting, though it names no party bound by the restriction and no consequence 4. The head detection and face recognition note does not use the words biometric or consent, and states no opt-out, no un-enrolment procedure, and no location or lifetime for an enrolled face representation 6.

Regarding demographic parity, the head detection and face recognition note publishes no accuracy figure and none disaggregated by population, while the gesture recognition note beside it claims parity across demographics and names skin tone 5.

See also

See also Cisco AI transparency technical notes. See also The Cisco Trust Portal as a monitorable document source. See also Named open model checkpoints and third party vendors behind Cisco AI features. See also Cisco biometric AI features and their governance. See also Webex Calling AI Receptionist. See also AI in Cisco Secure Email Threat Defense. See also Employment decision safeguards on automated agent scoring.

Applicability

Across the claims in this article, the evidence covers Cisco AI features across cloud deployments, verified as of 2026-09-05.

What remains uncertain

Whether and how consent is obtained for face recognition is not covered by any claim in this article. Where an enrolled face representation is stored and for how long is not covered by any claim in this article. The accuracy of head detection and face recognition across populations is not covered by any claim in this article. The 32 AI transparency notes not yet read are not covered by any claim in this article. Which specific named open model checkpoints Cisco builds on is not covered by any claim in this article. The specific details of the AI Receptionist 7 region residency table and its footnote are not covered by any claim in this article. Who is bound by a transparency note restriction and what follows a breach is not covered by any claim in this article. Why the employment restriction appears only in contact centre notes is not covered by any claim in this article.

WarmTransfer. (2026, September 5). The Cisco AI transparency note set beyond the contact centre. WarmTransfer. https://warmtransfer.net/knowledge/cisco-ai-transparency-census

@misc{warmtransfer-cisco-ai-transparency-census,
  title  = {The Cisco AI transparency note set beyond the contact centre},
  author = {{WarmTransfer}},
  year   = {2026},
  url    = {https://warmtransfer.net/knowledge/cisco-ai-transparency-census},
  note   = {Verified 2026-09-05}
}