Cisco states that Topic Analytics has a toxic hallucination filter which prevents the model generating new toxic harmful or unsafe language, and that toxic language already present in the source transcript may still surface because the filter does not alter the input.

fact · verified 2026-09-05

if toxic, harmful, or unsafe language is present in the original transcript, it may still surface in the transcript and call driver results, as the filter does not alter the source input.
Vendor
Cisco
Product
Webex Contact Center
Subsystem
AI transparency and model disclosure
Deployment
cloud
Region
not restricted
Release range
current as of 2026-09-05
Status
fact
Confidence
0.90
Verified
2026-09-05

Sources

Cite this claim

APA

WarmTransfer. (2026, September 5). Cisco AI transparency technical notes: claim wxai-tn-topic-toxic-filter-does-not-cover-source. WarmTransfer. https://warmtransfer.net/knowledge/claims/wxai-tn-topic-toxic-filter-does-not-cover-source

BibTeX

@misc{warmtransfer-claim-wxai-tn-topic-toxic-filter-does-not-cover-source,
  title  = {Cisco AI transparency technical notes: claim wxai-tn-topic-toxic-filter-does-not-cover-source},
  author = {{WarmTransfer}},
  year   = {2026},
  url    = {https://warmtransfer.net/knowledge/claims/wxai-tn-topic-toxic-filter-does-not-cover-source},
  note   = {Claim wxai-tn-topic-toxic-filter-does-not-cover-source, verified 2026-09-05}
}

Read this claim in context