Fiddler AI page on open-weight model PII leakage is unavailable

Fiddler AI’s page titled “Evaluating Open-Weight Models for PII Leakage in Customer-Service Agents” is currently unavailable. The page displays a not-found message instead of the article’s research or findings.
- The page is titled “Evaluating Open-Weight Models for PII Leakage in Customer-Service Agents.”
- The supplied URL currently displays a not-found message.
- The source provides no evaluation results, model list, methodology, or recommendations.
- Fiddler AI’s site describes its Control Plane as “The System of Trust for the Agentic Lifecycle.”
The listed topic
The page title identifies an evaluation of open-weight models for personally identifiable information, or PII, leakage in customer-service agents. The supplied page does not provide details about the models tested, the evaluation method, the results, or any recommendations.
What the page shows
Instead of the article, the page says, “We can't find the page you're looking for. That link didn't work.” It suggests learning more about Fiddler AI or contacting the company, and includes a link to go home.
Why the missing details matter
Without the article content, readers cannot verify which open-weight models were assessed or determine how Fiddler AI characterized their PII leakage risks. The page’s title indicates a focus on privacy in customer-service agents, but the available source does not support further claims about safety, performance, or model rankings.
Fiddler AI’s surrounding site navigation describes its Fiddler AI Control Plane as “The System of Trust for the Agentic Lifecycle.” It also lists continuous evaluations, agentic observability, policy enforcement, and AI governance, risk management, and compliance among its offerings. Those descriptions provide company context, but they do not replace the missing research page.
Fiddler AI: Evaluating Open-Weight Models for PII Leakage in Customer-Service Agents


