AI Content Notice

1. How models are used

We use language models to locate and extract information in long documents and to draft summaries. A model is an extraction method, never a source. Every extracted fact is designed to point to a source passage, and extractions are checked against structured data where possible.

2. Public content

Text on this site is generated automatically by a language model from public filings. GDLG does not review every item before publication. Every factual statement is linked to the cited passage. Facts with weak evidence are labelled Unverified. The public site does not publish severity, red flags, ratings, or investment advice.

3. Private deployment

In the Workstation, language models run on the customer's own hardware. Every factual sentence is designed to link to a source passage, and sentences that are inferences are labelled as such. Customer queries and documents do not leave the machine.

4. Limits

Language models can make mistakes, including misreading a source. We reduce that risk with quote verification, source and accession checks, schema and value checks, banned language checks, and an Unverified label for weak evidence. We do not claim that every item is correct. Always check the cited passage.

5. Training

We do not use customers' queries or data to train models. The private architecture is designed so that customer content never reaches our systems.