elloe ai: The AI Immune System - Disrupt 2025

Elloe AI: Building an "Immune System" for Artificial Intelligence

Owen Sakawa, the founder of Elloe AI, envisions his platform functioning as a crucial safeguard for AI systems – an “immune system for AI” and an “antivirus for any AI agent.”

Sakawa detailed his concept in a recent interview leading up to the TechCrunch Disrupt conference, where Elloe AI has been selected as a Top 20 finalist in the Startup Battlefield competition.

Addressing the Risks of Rapid AI Development

The core idea is to integrate a protective layer into companies’ Large Language Models (LLMs). This layer would proactively identify and mitigate issues like bias, inaccuracies (hallucinations), errors, non-compliance, the spread of misinformation, and the generation of unsafe content.

“AI is progressing at an unprecedented rate, yet this advancement occurs without sufficient safety measures, protective mechanisms, or systems to prevent potential failures,” Sakawa explained.

How Elloe AI Functions

Elloe AI is offered as an API or SDK – a module designed to sit atop an AI model’s output layer. Sakawa describes it as an “infrastructure on top of your LLM pipeline.”

Essentially, the system meticulously fact-checks every response generated by the underlying AI model.

A Multi-Layered Approach to AI Safety

The startup’s system employs multiple layers, referred to as “anchors,” to ensure comprehensive safety.

• The initial anchor verifies the LLM’s response against established, verifiable sources.
• The second anchor assesses the output for potential violations of regulations, including laws like HIPAA (U.S. health privacy) and GDPR (European data protection). It also identifies potential exposure of Personally Identifiable Information (PII).
• The final anchor creates a detailed audit trail, documenting the decision-making process. This allows regulators and auditors to analyze the model’s reasoning, pinpoint the sources of information, and evaluate the confidence levels associated with each decision.

Distinguishing Elloe AI from LLM-Based Solutions

Sakawa emphasizes that Elloe AI is not itself built upon an LLM. He believes that using LLMs to monitor other LLMs is akin to treating a problem with a similar source, offering only a temporary fix.

However, the system does leverage AI techniques, including machine learning. Human oversight is also integral, with Elloe AI employees continuously monitoring evolving regulations related to data and user protection.

Interested in learning more about Elloe AI and exploring other innovative companies? Join us at Disrupt, October 27 to 29, in San Francisco. Learn more here.