Yash BarveYash Barve

Position Paper: Defense of Proprietary AI-Trading Algorithms Against Mandatory Disclosure

This paper argues that organizations should be allowed to keep proprietary AI-designed trading algorithms private rather than disclose them publicly. Public disclosure could increase market instability by making trading behavior predictable and easier to exploit, while also reducing incentives for firms to innovate. Instead, the paper proposes confidential regulatory transparency, where algorithms are shared only with regulators, allowing oversight without sacrificing intellectual property protection. Note: This paper was written for CS492, a CS ethics course at the University of Waterloo.
March 12th, 2026
Image credit: Built In

Introduction

This position paper details our arguments as Team B for RPE 4. Specifically, we argue that organizations should retain the right to maintain privacy over proprietary AI-designed trading algorithms. Industries such as financial trading rely heavily on the information and data that organizations can collect and sort, especially algorithms that they design using said data. Thus, we believe that algorithms designed by AI are no less deserving of protective rights than those designed by people.

Proprietary Algorithms Are Trade Secrets

First, we believe that proprietary algorithms should remain as organizational trade secrets. In a document prepared by the World Intellectual Property Organization (WIPO), it is stated that for some information to qualify as a trade secret it must be:

  • “commercially valuable because it is a secret”
  • “be known only to a limited group of persons”
  • “be subject to reasonable steps taken by the rightful holder of the information to keep it a secret, including the use of confidentiality agreements for business partners and employees”
    (WIPO, n.d.)

Additionally, it is stated that:

“Trade secrets encompass both technical information, such as information concerning manufacturing processes, experimental research data, software algorithms and commercial information…” (WIPO, n.d.)

It is clear to see that proprietary trading algorithms fit the category of trade secrets — they are valuable, are known only by employees of an organization, and are technical information. Regardless of whether they are AI-generated or not, organizations are not required to make intellectual property such as trading algorithms publicly available.

For further proof that trading algorithms are protected information, there was a court case between a bank and an employee for stealing and replicating their high-frequency trading code. In United States v. Agrawal (2013), a former employee of a bank was arrested for stealing code for their high-frequency trading application considered proprietary information.

From the court case and the WIPO definition, it is clear to see that all intellectual property, including trading algorithms, are protected under law regardless of the advantage that they provide an organization. We believe that algorithms created by artificial intelligence deserve the same level of protection as shown to algorithms developed by hand.

Unique Algorithms Prevent Market Instability

Beyond their legal protection, proprietary AI trading algorithms play an important functional role in preserving systemic resilience within financial markets. While critics argue that automated systems can amplify volatility, empirical research suggests that algorithmic trading frequently improves liquidity and enhances price discovery under normal conditions (Hendershott et al., 2011).

Firms develop these systems independently using different datasets, model architectures, execution speeds, and risk tolerances.

This variation in strategy reduces the probability that all market participants will react identically to the same signal at the same time. Financial markets operate as complex adaptive systems, and stability within such systems depends on decentralized decision-making rather than uniform behavior.

Mandating public disclosure of proprietary algorithmic logic would likely increase instability by making trading behavior more predictable and easier to exploit. In adversarial environments, predictability can invite strategic manipulation, front-running, and coordinated responses that amplify price movements (Budish et al., 2015).

If firms are required to reveal the internal triggers or parameters guiding their systems, other actors may design strategies specifically to exploit those known mechanisms. Rather than reducing volatility, such exposure may increase systemic fragility by accelerating reactive trading dynamics.

For these reasons, preserving proprietary independence while maintaining structural safeguards represents a more prudent approach to market stability than compulsory public disclosure.

Inequality Creates Innovation

Proprietary algorithmic trading systems represent the output of enormous capital investment that goes into data infrastructure, hiring the best talent, and iterative development.

Organizations undertake this investment precisely because it offers a competitive advantage in return.

If mandatory disclosure eliminates that exclusivity, the rational response for any organization is to invest less. Boldrin and Levine (2013) argue that intellectual property protections are mechanisms to ensure that innovators can recover the costs of creation.

The result of forcing disclosure of trade secrets would not create fairness, but would reduce innovation and long-term market efficiency. The development of sophisticated tools and advanced algorithms would slow down considerably due to the absence of any recoverable return on that investment.

This concern reflects the foundational logic behind trade secret and intellectual property law more broadly. The legal system grants organizations protection over proprietary innovations specifically because, without that protection, the incentive to develop them collapses (WIPO, n.d.).

Forcing the disclosure of AI-designed trading algorithms would effectively convert years of private R&D into a public good at the organization’s expense.

History supports this: industries that have enjoyed strong IP protections have consistently produced greater rates of technological advancements than those that have not (Lerner, 2009). Weakening the protections in algorithmic trading would lead to convergence toward simpler and less ambitious strategies.

Alternative Solution: Algorithmic Transparency

We understand that the concerns driving calls for mandatory disclosure are legitimate. However, it is important to distinguish between regulatory transparency and public transparency. These are not the same thing, and conflating them leads to policy responses that are far more invasive than necessary.

Our position is not that algorithmic trading should operate without any oversight, but that the correct approach is confidential regulatory transparency.

Organizations should be required to submit their algorithmic logic, parameters, research documentation, and any relevant data directly to relevant regulatory bodies, such as the SEC and CFTC in the United States, or the OSC in Canada.

These bodies already enforce Regulation SCI, market access rules, and anti-manipulation provisions that give regulators meaningful supervisory power (SEC, 2014; Ontario Securities Act, 1990).

There should be binding confidentiality agreements in place to prevent any misuse or transfer of that information. This way regulators gain access needed to identify systemic risks. At the same time, firms retain protection over the competitive value of their proprietary systems.

This approach is already the standard in other industries.

For example:

  • Pharmaceutical companies submit detailed proprietary formulations and trial data to the FDA without disclosing them publicly.
  • The European Union's AI Act emphasizes documentation, auditability, and accountability rather than forced source code disclosure (European Parliament & Council of the European Union, 2024).

If this approach is applied to algorithmic trading, it would allow regulators to conduct detailed post-event analysis. A clear example is the 2010 Flash Crash, where the SEC and CFTC examined firm-level trading data and implemented structural safeguards without requiring public exposure of proprietary systems (SEC & CFTC, 2010).

Confidential regulatory access achieves every legitimate oversight goal while preserving the intellectual property protections and R&D incentives that make algorithmic trading innovation possible.

Conclusion

To conclude, there are numerous reasons that algorithms designed by organizations remain private, and AI-generated algorithms deserve no less protection.

The WIPO framework and the United States v. Agrawal case provide clear evidence that algorithms are protected intellectual property. There is no reason why AI trading algorithms should not be considered intellectual property as well.

Maintaining privacy surrounding algorithms:

  • helps preserve market stability
  • prevents manipulation that could arise from predictable strategies
  • encourages innovation through competition and investment

Additionally, regulatory transparency can still ensure oversight without forcing public disclosure.

By revealing algorithms only to appropriate regulatory boards, firms can ensure they operate within industry rules while maintaining the advantages of proprietary privacy that sustain innovation and competition.

Works Cited

Collaborators

  1. Yash Barve – yhbarve@uwaterloo.ca
  2. Nicholas Rebello - nrebello@uwaterloo.ca
  3. Roman Guevarra - raguevar@uwaterloo.ca