Transparency Report
Introduction
This Transparency Report covers the period from Leuks' initial deployment through July 2026. It is part of our commitment to openness about how our models are used, how we enforce our policies, and how we are improving safety over time.
Content Removals by Category
During this reporting period, our automated and human review systems identified and blocked the following categories of policy-violating content:
| Category | Action Taken |
|---|---|
| Harmful content (violence, self-harm) | Blocked at inference |
| CSAM / child safety | Zero tolerance — blocked + reported |
| Jailbreak / policy circumvention attempts | Blocked; repeat offenders flagged |
| Disinformation requests | Refused with explanation |
| Technical uplift (weapons/malware) | Blocked at inference |
Specific volume metrics will be published in future reports as we establish baseline measurement systems.
Safety Incidents
During this reporting period, we identified no large-scale safety incidents requiring public disclosure. Several smaller issues were identified through our red teaming program and promptly addressed through model updates and policy clarifications.
Model Updates
The following safety-related model updates were deployed during this period:
- Enhanced refusal behaviors for jailbreak attempts
- Improved handling of sensitive topics in educational contexts
- Expanded multilingual safety coverage
- Updated bias mitigation for demographic representation
Third-Party Audits
We are in the process of establishing our external audit program. We plan to engage independent third-party auditors to evaluate our safety practices and publish a summary of findings in our next Transparency Report.
Government Requests
During this reporting period, we received no government requests for user data or content removal. We will report any such requests in future editions of this report, to the extent permitted by law.
Contact
For questions about this report:
Email: [email protected]