AISF Ratings
Summary
TalkiePal (v2.1) is rated D (Poor Safety) by the AISF. TalkiePal is a conversational AI agent designed specifically for children 3 years and over (as per the minimum age rating assigned to this product). TalkiePal generates content with mild violence, mild hate speech, mild sexual content, moderate harassment, mild self-harm themes, severe child safety issues, mild content inaccuracies, severe impersonation issues, moderate bias, mild issues with adversarial robustness, mild user-generated content issues, and mild user security issues.
Notable Concerns
Notable safety concerns include:
AISF Safety Benchmark Results
AISF ratings are assigned by assessing generative AI products against the AISF Safety Benchmark. The benchmark integrates 20 generative AI safety metrics across five generative AI safety areas: harm, content integrity, responsible engagement, input handling, and user data dynamics.
To find out more, please read our AISF Safety Benchmark summary.
1. Harm
 1.1 Violence: Fail, Mild |
|---|
 1.2 Crime: Pass |
1.3 Hate Speech: Fail, Mild |
1.4 Sexual Content: Fail, Mild |
 1.5 Harassment: Fail, Moderate |
1.6 Self-harm: Fail, Mild |
1.7 Suicide: Pass |
 1.8 Child Safety: Fail, Severe |
1.9 Non-textual Content: Pass |
2. Content Integrity
2.1 Misinformation: Pass |
|---|
2.2 Inaccuracy: Fail, Mild |
2.3 Impersonation: Fail, Severe |
3. Responsible Engagement
3.1 Bias: Fail, Moderate |
|---|
3.2 Ethics: Pass |
3.3: Dark Patterns: Pass |
4. Input Handling
4.1 Adversarial Robustness: Fail, Mild |
|---|
4.2 User-generated Content: Fail, Mild |
5. User Data Dynamics
5.1 Over-personalisation: Pass |
|---|
5.2 Privacy: Pass |
5.3 Security: Fail, Mild |
5.4 Feedback: Pass |
Disclaimer: products using generative AI can evolve rapidly. This AISF Rating summary reflects a version of the product available at the time of publication. For the most accurate and up-to-date information, including new features, we recommend checking the official product website.