Patent filing for unique AI-powered algorithm for identifying top-performing AI and human decision-makers, including which LLM is most accurate for a given task
System identifies the top 1–5% of AI models and human experts per decision type, applying patent-pending bias-removal across financial trading, clinical decision support, fraud detection, cybersecurity, and additional decision-intensive domains

Perpetuals.com Ltd (NASDAQ:PDC), an AI-powered fintech company, has filed a U.S. patent application covering BayesShield AI, a system that identifies which AI model or human expert performs best for a specific decision, in real time. The patent addresses one of the most-cited blockers to enterprise agentic AI deployment: the gap between average benchmark performance and reliable performance on the decision in front of you.

The trailblazing and complex system is built on a simple premise: performance is not universal. Both AI models and human experts perform differently depending on the context of a decision. A model that performs well in one type of environment may underperform in another, and the same is true for human decision-makers.

Perpetuals's new framework addresses this by continuously evaluating performance across different decision environments, or "regimes," and learning which AI models or human experts are most effective in each. It can assess leading large language models, including ChatGPT, Claude, Gemini, and DeepSeek systems, on a task-by-task basis rather than relying on a single overall benchmark.

The patent application addresses what industry analysts have identified as the central blocker to enterprise agentic AI deployment. McKinsey's 2026 AI Trust Maturity Survey found that security and risk concerns are now the top barrier to scaling agentic AI, with only about one-third of organizations reaching maturity in agentic AI controls. Gartner has projected that more than 40% of agentic AI projects will be canceled by 2027 due to escalating costs, unclear business value, or inadequate risk controls. Recent academic research has documented a 37% gap between lab benchmark scores and real-world production performance, the gap that BayesShield is designed to close.

Over time, the system builds a map of which models and experts are best suited to specific conditions and routes decisions to them only when those conditions are present. This enables more precise matching between problem type and decision-maker, rather than relying on averages or static rankings.

"For a decade, AI progress has been measured by which model has the best benchmark average. The agentic era has exposed that this metric doesn't predict production performance; McKinsey, Gartner, and Deloitte are all reporting the same gap. The next edge in AI isn't a single model that wins on average. It's a system that knows which model, or which expert, will win in the decision in front of you," said Patrick Gruhn, CEO of Perpetuals. "That's what we are patenting. And it's already running our products today."

The approach is designed for use across a range of decision-intensive industries, including financial trading, fraud detection, cybersecurity, content moderation, health care diagnostic decision support, insurance underwriting, and legal analysis. This process is already powering many of Perpetuals's own products, including in financial markets and the deployment of BayesShield to reduce medical diagnostic errors, with more on the way.

The patent application describes a six-step pipeline:

Capturing predictions in a simulated, no-risk environment
Tagging each prediction with the conditions in which it was made
Training AI models on rare but high-value cases rather than discarding them
Identifying the top 1–5% of performers in each situation type
Deploying these top performers only when conditions match their proven strengths
Continuously updating rankings with new performance data

About Perpetuals.com Ltd.

Perpetuals.com Ltd (NASDAQ:PDC) is a fintech company developing AI-powered trading products and prediction markets, with a global footprint across the United States, Europe, and Asia. Its mission is to reduce risk through empowering retail users with intuitive, secure, and efficient trading experiences across multiple asset classes.

Perpetuals's proprietary trading platform, Kronos X, combines advanced AI and data analysis. The technology is trained on billions of trades, monitors market activity in real time, identifies patterns for trading and risk decisions, and provides multi-asset coverage with self‑clearing blockchain-based settlement. The company's licensed European Multilateral Trading Facility (MTF) infrastructure and Kronos X multi‑asset exchange platform operate with full MiFID II, MiCA, DORA, and EMIR compliance.

Forward-Looking Statements: This press release contains forward-looking statements as defined within the meaning of the Private Securities Litigation Reform Act of 1995, including, but not limited to, statements regarding the scope and protection afforded by the provisional patent application, the expected applicability of the disclosed methodology across industries, and the anticipated benefits to the company's intellectual property portfolio. The filing of a provisional patent application does not guarantee that any patent will issue, that any claims will be granted as filed, or that any resulting patent will provide meaningful commercial protection. Words such as "expect," "will," "positions," "advancing," "projected," "anticipated," and other similar expressions indicate forward-looking statements, though not all forward-looking statements contain such words. These statements reflect the company's current view with respect to future events, are subject to risks and uncertainties that could cause actual results to differ materially, including regulatory approvals, market conditions, the ability to realize anticipated benefits of any patent that may be issued, and risks detailed in the company's filings with the Securities and Exchange Commission, and are necessarily based upon a number of estimates and assumptions that, while considered reasonable by the company, are inherently subject to significant business, economic, competitive, regulatory, political, and social uncertainties and contingencies. In addition, there can be no assurance that any patent will issue from the company's pending application, or that any issued patent will provide protection of sufficient scope to provide meaningful competitive advantages. Should one or more of these risks or uncertainties materialize, or should the assumptions set out by the company underlying those forward-looking statements prove incorrect, actual results may vary materially from those described herein. Individuals are cautioned that forward-looking statements are not guarantees of future performance and accordingly investors are cautioned not to put undue reliance on forward-looking statements due to the inherent uncertainty therein. These forward-looking statements are made as of the date of this press release and the company does not undertake any obligation to update these forward-looking statements, except as required by law.

View source version on businesswire.com: https://www.businesswire.com/news/home/20260507266161/en/

Media

Perpetuals Email: press@perpetuals.com Website: www.perpetuals.com

Investor Relations

ir@perpetuals.com

Perpetuals Files Patent for Bias-Removal Architecture for AI Agents Solving The Reliability Problem Blocking Enterprise Agentic AI