A website called In the Weights now lets you check whether AI language models have encoded knowledge about you during training, assigning a strength score based on how prominently you appear in their weights.

What happened: Joey Flynn and Thomas Dimson, both former OpenAI employees, built In the Weights, a site that queries several language models to determine who a specific person is and assigns them a strength score. The maximum strength score is 996, reserved for names like Mozart, Shakespeare, or Taylor Swift. The site's creators note that appearing in smaller models like Meta's Llama (which has a billion parameters) counts as being highly relevant.

Why it matters: The site reveals which people were considered important enough during a model's training for it to encode them into its weights—the billions of numerical values where AI models store their knowledge. If you show up, the model learned about you well enough to recall information without needing external tools like web search. This offers a tangible way for individuals to understand their footprint in AI training data.

What to watch: The creators flag important limits: models can hallucinate biographical details, typos reduce scores, and common names often produce worse results. These caveats mean a score should not be read as a definitive measure of how well a model truly knows someone.