# Telluvian

> Telluvian is a real-time AI hallucination detection system powered by mechanistic interpretability. It inspects a model's internal activations as it generates a response and returns a confidence signal alongside the output, so users can tell which parts of an AI-generated answer to trust. Telluvian is based in London and is part of the London Initiative For Safe AI.

Telluvian works by submitting a query to a frontier model, running probing classifiers over that model's internal activations while the response is generated, and returning the response together with a token-level trust indicator. For models Telluvian does not have direct weight access to, it uses an approach called Proxy White-Box Interpretability: measuring internal directions (such as refusal or uncertainty) in an open-weight model with a similar training distribution, and using that as an implicit proxy measurement inside the closed model.

Primary use cases: legal teams (catching hallucinated case citations, misquoted statutes, and fabricated facts in AI-assisted drafting and research), finance and compliance teams (verifying AI outputs in regulated workflows), and AI product teams (integrating Telluvian as a verification layer to give end users confidence signals on every output).

## Key pages

- [Homepage](https://telluvian.ai/): Overview of Telluvian, how it works, and who it's for.
- [Careers](https://telluvian.ai/careers): Open roles at Telluvian.
- [Contact](https://telluvian.ai/contact): Get in touch about the product, enterprise pricing, or careers.

## Blog: Research

- [Why AI Companies Will Have to Embrace Manifold Learning](https://telluvian.ai/blog/manifold-learning): Why manifold learning matters for AI companies, from the manifold hypothesis to intrinsic dimensionality in transformer hidden states.
- [The Two-Neighbour Method: How to Measure a Dataset's "True" Dimensions](https://telluvian.ai/blog/twonn): How many intrinsic dimensions does your data really have? An introduction to twoNN and what it reveals about a transformer's residual stream.
- [What is Mechanistic Interpretability?](https://telluvian.ai/blog/what-is-interp): Modern large language models are immense black boxes, yet they still represent internal features in a standard way we can begin to understand.

## Blog: Engineering

- [Is Your AI Hallucinating?](https://telluvian.ai/blog/is-your-ai-hallucinating): How our white-box proxy model gives you a per-token hallucination score, including exactly what it costs.
- [AI Portal](https://telluvian.ai/blog/ai-portal): Why AI inference costs keep rising despite falling token prices, and how mechanistic interpretability techniques like routing and prompt compression can cut them without sacrificing quality.
- [How We Detect AI Hallucinations in Real-Time](https://telluvian.ai/blog/hallucination-detection-explained): The technical approach behind Telluvian's real-time hallucination detection, including probing classifiers, activation analysis, and token-level confidence scoring.

## Blog: Opinion

- [AI Safety Needs Startups](https://telluvian.ai/blog/ai-safety-needs-startups): Why the best way to deploy safety at scale might be to sell it.

## Contact

- Email: hello@telluvian.ai
- Address: London Initiative For Safe AI, 25 Holywell Row, London EC2A 4XE, United Kingdom

## Full index

- [Blog index](https://telluvian.ai/blog): All posts, each tagged Research, Engineering, or Opinion.
- [Sitemap](https://telluvian.ai/sitemap.xml): Complete list of indexable URLs.