Clarifying DeepIP’s Compliance With AI Codes of Conduct

Many patent practitioners are wary about trusting generative artificial intelligence (AI) products for patent drafting. It’s natural for legal professionals to wait before adopting these platforms due to shifting frameworks for ethical codes of conduct related to generative AI.

‍

DeepIP believes in addressing these frameworks directly, which has resulted in strong relationships with several major corporate customers. These companies, which are well practiced at performing due diligence, often find that DeepIP’s confidentiality and security features exceed their expectations.

‍

Unlike many other generative AI providers, DeepIP also works with patent attorneys to help them understand their own responsibilities in adopting its AI patent assistant. We don’t just sell a service; we empower legal professionals to improve their productivity with the confidence that they can meet their obligations both to their clients and the entire legal profession.

‍

Below is a quick overview of recent developments to patent attorney ethical frameworks in response to generative AI. We explain how DeepIP satisfies the technical concerns of those frameworks, and any additional obligations that patent attorneys must satisfy.

‍

Institute of Professional Representatives before the European Patent Office (EPI)

‍

In November 2024, the Professional Conduct Committee of EPI, which provides legal opinions on European patent attorney practice, issued guidelines on the use of generative AI to draft patent applications and respond to office correspondence. These guidelines separate generative AI applications, which produce text and image outputs, from non-generative AI, such as spelling checkers and grammar suggestion tools.

‍

EPI’s guidelines identify overarching principles including high standards for honesty and moral decency, maintaining confidentiality, and putting client interests first as required by EPI’s Code of Conduct. These principles are clearly spelled out over eight separate guidelines.

‍

Evaluating DeepIP’s Ethical Compliance in Generative AI Usage

‍

Guideline 1: Members should inform themselves of general characteristics and specific attributes of generative AI models, including prompt confidentiality and the likelihood of hallucinations.

• DeepIP’s Advantage: Our security page identifies our compliance with international standards for data security and information management. We have implemented some of the strictest cybersecurity standards including ISO270001 and SOC2 Type II, and we are compliant with both the EU’s GDPR and HIPAA in the U.S.
• Practitioner Obligations: Every generative AI platform has the potential to produce hallucinations, which can be addressed through close proofreading of all outputs.

‍

Guideline 2a: Members must ensure adequate confidentiality of training datasets, instruction prompts and other content transmitted to AI models. Members should not use AI models if any doubt exists that appropriate levels of confidentiality can be maintained.

• DeepIP’s Advantage: We operate our platform through a zero-retention application programming interface (API), which prevents any data created by customer inputs from leaking into AI training models. DeepIP protects stored data according to ultra-secure AES-256 encryption and employs TLS 1.2+ cryptography protocols for protecting data in transit.
• Practitioner Obligations: Communicating data security needs specific to your practice area will help DeepIP develop personalized solutions for putting our AI platform to work for you.

‍

Guideline 2b: Members must inform themselves of the likelihood of disclosures deriving from the use of specific AI models.

• DeepIP’s Advantage: Along with our zero-retention API, we employ technical controls for continuous data monitoring meeting governance standards promulgated by agencies like NIST.
• Practitioner Obligations: Look for generative AI vendors that are transparent about their cybersecurity standards and responsive to customer concerns.

‍

Guideline 3a: Members cannot cite the use of generative AI as an excuse for errors or omissions.

• DeepIP’s Advantage: We are highly responsive to client needs and provide industry-leading tools for minimizing generative AI errors in the patent prosecution context.
• Practitioner Obligations: Just as model rules of legal ethics give supervising attorneys responsibility over their subordinates, patent attorneys using generative AI solutions must bear the ultimate burden of ensuring the accuracy of their filings.

‍

Guideline 3b: Members must check their work product to ensure that it is as if the work product was produced by a competent human practitioner.

• DeepIP’s Advantage: Our patent drafting solution identifies missing formalities and other omissions that must be addressed prior to filing. We also highlight errors in claim language or figure labels to minimize mistakes.
• Practitioner Obligations: Ethical obligations for legal practitioners require close proofreading to ensure the accuracy of documents before signing and filing them, regardless of the use of generative AI systems.

‍

Guideline 4: Members must establish client wishes on the use of generative AI before starting such use.

• Practitioner Obligations: Patent attorneys have the same duties of client communications that all legal professionals must observe. Before subscribing to DeepIP, obtain clear confirmation, ideally in a signed writing, that your clients are comfortable with your use of an industry-leading generative AI application for patent drafting and prosecution.

‍

Guideline 5a: Members may advertise their use of generative AI tools, but such statements must be accurate and not promote discrimination between members.

• Practitioner Obligations: Statements about the use of DeepIP, including customer testimonials, are fine, but those communications cannot disparage the practice of other patent attorneys for using other generative AI tools or avoiding AI altogether.

‍

Guideline 5b: Members are not required to disclose their use of generative AI, unless binding statutes, rules, orders, or client instructions otherwise oblige them.

• Practitioner Obligations: There is currently no general duty to disclose generative AI use to the European Patent Office or the Unified Patent Court, although patent attorneys may have to follow specific orders to do so. Like attorney advertising, such statements must be accurate and may not disparage other attorneys.

‍

Guideline 6: Members must establish independent user accounts for each client if the nature of the AI model employed creates a risk that training prompts related to one client could be transferred to the work of another client.

• DeepIP’s Advantage: Data confidentiality is paramount to us. We do not retain any client data or use user-provided content for training our models. All data in transit is encrypted, and we maintain complete segregation across different users and organizations.
• Practitioner Obligations: Patent attorneys should prioritize generative AI platforms that demonstrate the strongest commitment to data security and confidentiality. Key considerations include data encryption, APIs with no data retention, strict data segregation with controlled access rights, and a guarantee that data will not be reused for training purposes.

‍

Guideline 7a: Members must be aware of relevant legislation impacting generative AI use, and must comply with their relevant provisions.

• Practitioner Obligations: As the EPI guidelines point out, several provisions of the EU’s Artificial Intelligence Act are effective as of January 2025. Patent attorneys operating in the EU must respect legal developments while also following these guidelines.

‍

Guideline 7b: Members should also respect requirements imposed by professional organizations and regulators that may impact their generative AI use.

• Practitioner Obligations: Many bar associations and even indemnity insurance providers are implementing rules on generative AI use that patent attorneys must follow in the normal course of business.

‍‍

Guideline 8: Members can only charge fees for AI-generated work product that fairly reflect the time and effort, degree of difficulty and/or the degree of risk involved. Those fees can include charges for training AI tools, AI subscription fees or checking AI-generated work.

• ‍Practitioner Obligations: Similar to advertising statements, while patent attorneys are permitted to charge for the use of generative AI-related services, such charges must be reasonable. This also underscores that patent attorneys can include time spent checking AI-generated work product in their billable hours.

‍

If you have any further inquiries about DeepIP’s adherence to EPI guidelines, please don't hesitate to reach out to us!