Provision record
Cohere · Cohere Responsible Use Policy · View original document ↗

Safeguards in place against harmful text generation

Medium severity High confidence Explicit document language Common · 141 of 352 platforms

Key Facts

What has Cohere put in place to avoid generating harmful text?
Cohere states that it has put safeguards in place to avoid generating harmful text, but that it is still possible to encounter toxicity, especially over long conversations with multiple turns.
Stay ahead of the changes
Track Cohere and get the diff the day its terms change.
Share 𝕏 Share in Share 🔒 PDF

This analysis describes what Cohere's agreement states, permits, or reserves. It does not constitute a legal determination about enforceability. Regulatory applicability and practical outcomes may vary by jurisdiction, enforcement context, and individual circumstances. Read our methodology

ConductAtlas Analysis

Why it matters (compliance & governance perspective)

Cohere's acknowledgment that toxicity remains possible despite safeguards is consequential for users and developers who may encounter harmful outputs during extended interactions.

Clause Stability Stable

0
Changes
3
Months Monitored
Jul 10, 2026
First Seen
Jul 10, 2026
Last Seen
This clause type exists across 350 other provisions on other platforms.

Consumer impact (what this means for users)

Users are informed that harmful or toxic text may still be generated by Cohere's model, particularly during long, multi-turn conversations, despite existing safeguards.

How other platforms handle this

Tinder Medium

Our systems and processes are generally designed to prioritize the detection and removal of the most serious and harmful violations (e.g., child sexual exploitation material or terrorist content)...

Microsoft Azure Medium

Microsoft uses the resulting text data to provide captioning of chat for players who need it. This data may also be used to provide a safe gaming environment and enforce the Community Standards for Xb...

OpenAI Medium

We start by teaching our AI right from wrong, filtering harmful content and responding with empathy.

See all platforms with this clause type →
▸ View Original Clause Language DOCUMENT RECORD
"
We have put safeguards in place to avoid generating harmful text, and while they are effective...it is still possible to encounter toxicity, especially over long conversations with multiple turns.

Excerpt from Cohere's Responsible Use Policy

Applicable regulations

California AB 2013 AI Training Data Transparency
US-CA

Provision details

Document information
Document
Cohere Responsible Use Policy
Entity
Cohere
Document last updated
May 12, 2026
Tracking information
First tracked
July 9, 2026
Last verified
July 9, 2026
Record ID
CA-P-064314
Document ID
CA-D-00830
Evidence Provenance
Source URL
Wayback Machine
Content hash (SHA-256)
06ab5e02539ada114dce05fa235995dca7762c7cb5d195703306b5ef9bf69fc6
Analysis generated
July 9, 2026 05:27 UTC
Methodology
Evidence
✓ Snapshot stored   ✓ Hash verified
Citation Record
Entity: Cohere
Document: Cohere Responsible Use Policy
Record ID: CA-P-064314
Captured: 2026-07-09 05:27:24 UTC
SHA-256: 06ab5e02539ada11…
URL: https://conductatlas.com/platform/cohere/cohere-responsible-use-policy/provision/CA-P-064314/safeguards-in-place-against-harmful-text-generation/
Accessed: Sept. 8, 2026
Permanent archival reference. Stable identifier suitable for legal filings, compliance documentation, and research citation.
Classification
Severity
Medium
Categories

Other risks in this policy

Get the research letter

Companies change their terms quietly. We read every version and catch what actually changed. One email a week on the changes that matter and what they mean.

Frequently Asked Questions

What does Cohere's Safeguards in place against harmful text generation clause do?

Cohere's acknowledgment that toxicity remains possible despite safeguards is consequential for users and developers who may encounter harmful outputs during extended interactions.

How does this clause affect you?

Users are informed that harmful or toxic text may still be generated by Cohere's model, particularly during long, multi-turn conversations, despite existing safeguards.

How many platforms have this type of clause?

ConductAtlas has identified this type of provision across 141 platforms. See the full comparison.

Is ConductAtlas affiliated with Cohere?

No. ConductAtlas is an independent monitoring service. We are not affiliated with, endorsed by, or sponsored by Cohere.