Credal Connects Company Data to LLMs with Enhanced Security Measures

Credal.ai, a startup backed by Y Combinator, has successfully raised $4.8 million in a seed round led by Spark Capital. This innovative company connects enterprise data with text-generating AI models hosted in the cloud.

Founded by Jack Fischer and Ravin Thambapillai—both alumni of Palantir—the duo formed their partnership through a shared passion for security and compliance. Ravin, a former Google engineer, transitioned into tech after earning a degree in philosophy, politics, and economics from Oxford.

“We recognized our unique backgrounds in enterprise data security and AI at Palantir positioned us to build a trustworthy AI data platform for enterprises,” Fischer shared in an email interview.

Initially, Fischer and Thambapillai envisioned a “decision-making assistant” leveraging large language models (LLMs) similar to ChatGPT to aid executives in strategic decision-making by reading documents and offering advice. However, their vision expanded to create a robust tool that seamlessly connects internal data to external LLMs.

Today, Credal can be utilized to develop AI-powered chatbots for diverse applications, including domain-specific knowledge. For instance, enterprises can use Credal to create a bot that addresses security inquiries regarding proprietary software, referencing the most current documentation.

Credal does not operate LLMs itself; instead, it bridges the gap between users’ inquiries (like “What’s the latest version of this software?”) and third-party LLM APIs such as OpenAI or Anthropic's. This platform acts as a “co-pilot,” integrating smoothly with existing applications like Slack.

To enhance efficiency, Credal automatically directs prompts to the most suitable LLM when multiple options are available, considering factors such as data sensitivity, cost, company policy, and technical capabilities. In some scenarios, it even utilizes multiple LLMs to complete a task, combining the strengths of models like Anthropic’s Claude and GPT-4 to organize company documents.

While several platforms, such as Unstructured, Deasie, and LlamaIndex, offer solutions for connecting company data to LLMs, Credal stands out by prioritizing compliance and security—at least according to Fischer's insights.

Credal proactively redacts and anonymizes data while warning users before sensitive information is sent to a public cloud-hosted LLM. It also maintains logs detailing which data has been shared with specific LLMs.

By default, data sent to Credal is retained for 30 days after account expiration, which may concern some companies. However, Fischer assures that administrators can delete data at any point.

Fischer also claims that Credal is among the few vendors registered under the Data Privacy Framework—an agreement governing U.S.-EU personal data transfers. This compliance has facilitated contracts with regulated European companies like Wise.

“IT departments seek visibility and control over AI usage within their organizations,” Fischer noted. “Credal delivers transparency in a standardized format across various LLM providers, granting precise data controls over user access to models and data … Unlike other AI systems, Credal mimics the permissions of the connected source systems, ensuring AI responses are derived only from relevant, accessible company documents.”

Since its launch in April, Credal has managed over a quarter million LLM interactions and processed around 100,000 corporate documents. Currently, it serves 11 clients, several of which have already signed six-figure contracts.

With the newly acquired seed funding, Credal plans to increase its workforce, which currently comprises five employees, and enhance the product to encompass additional data sources and advanced data retrieval capabilities, according to Fischer.

“The AI industry currently faces a disconnect between the overwhelming enthusiasm and the relatively limited number of companies utilizing LLMs to create tangible value,” Fischer added. “Credal is addressing this gap by deeply integrating with a select few exceptional enterprises, effectively solving their challenges from start to finish.”