Can AI development and the free, widely available ChatGPT[1] replace a lawyer’s knowledge and experience? I examine a ChatGPT-drafted loan between a limited liability company and its board member, comparing its information with legal requirements and identifying risks.
ChatGPT’s “advice” and analysis
The prompt “prepare a loan agreement between a board member and a limited liability company” generated:
Here is an example outline. Such agreements should be carefully tailored to the circumstances. Always consult a corporate lawyer before drafting and signing.
LOAN AGREEMENT NO. [number]
Concluded on [date] between [board member’s name], residing at [address], PESEL [number], the “Lender”, and [company name], registered office [address], KRS [number], NIP [number], REGON [number], represented by [name and position], the “Borrower”.
§ 1. Subject
1. The Lender lends [amount] (in words: [amount]) złoty to the Borrower.
2. The term is [period] from signing.
§ 2. Interest
The loan is interest-free.
§ 3. Repayment
1. The Borrower will repay the full amount by [date].
2. Repayment is in one instalment.
3. Payment will be transferred to the Lender’s account [number].
§ 4. Security
The Lender receives a third-party surety from the Borrower in the form [security, e.g. property or personal surety].
§ 5. Final provisions
1. Effective on signature by both parties.
2. Unregulated matters are governed by the Civil Code and relevant limited liability company law.
3. Two identical counterparts, one per party.
[Date and signatures]
Adapt the agreement to specific conditions and national law and carefully consider clauses, especially security. Always consult a lawyer before signing.
ChatGPT itself twice stresses professional review, legal adaptation and the template’s illustrative, general nature. Correct and fair as this warning is, it also limits potential provider liability for adverse consequences of the “advice”.
The draft gives only general principal terms: amount, repayment, interest and possible security. Its generality may make it insufficient. Proper clauses require detailed consideration and lawful regulation.
Every transaction requires comprehensive analysis of risks and conditions for validity and effectiveness. An available or generated template may not suffice even for a simple act.
A further prompt asking about risks and requirements generated:
A loan between a company and a board member involves risks and requirements for both. Consider:
Company risks:
- Conflict of interest: a manager borrowing from the same company may compromise objective board decisions.
- Insolvency: inability to repay may affect the company’s finances.
- Company image: favourable terms for the member may affect perceptions among shareholders, investors and the public.
Company requirements:
- Corporate approval may be required under the articles or law, for example by a general meeting or supervisory board.
- Transparency: clear, documented terms and compliance with procedures.
Board member risks:
- Financial pressure as debtor may affect independent management decisions, particularly if the company faces difficulties.
- Legal obligations: failure to fulfil financial duties may lead to legal action.
Board member requirements:
- Understand repayment, interest, penalty provisions and consequences of default.
- Be aware that corporate approval may be needed and prepared for refusal.
Always consult a corporate and financial lawyer. Careful preparation and documentation minimise risk and ensure compliance.
First, this contradicts the earlier draft: the company was borrower, but the risks assume it is lender, within the same chat. A non-lawyer might also conclude there are no special requirements, the template is sufficient, and approval is optional because limited liability companies often lack supervisory boards and do not have the joint-stock company’s general meeting mentioned. Apart from approval and conflicts, the identified risks were commercial rather than formal legal requirements.
ChatGPT also answers only questions asked and does not anticipate further issues. A non-lawyer may not realise additional analysis is needed.
Issues ChatGPT omitted
- Corporate approvals
The vague reference to possible corporate approval is insufficient.
Loans with board members face restrictions because directors represent the company and could otherwise contract without shareholders’ consent or knowledge.
Shareholder approval is necessary regardless of who lends. It must be expressed by resolution, generally minuted with strict formal requirements.
A subsidiary contracting with a parent’s board member, commercial proxy or liquidator may also require the parent’s shareholder approval.
Missing consent may invalidate the agreement.
The articles may require additional approvals, an issue entirely omitted.
- Company representation
Beyond naming a “representative”, ChatGPT ignored representation rules.
Normally the board represents under the articles, for example each member independently. For a loan with a board member, however, the company must act through its supervisory board or an attorney appointed by shareholder resolution.
Incorrect representation makes the agreement invalid, potentially requiring restitution, creating damages liability and making recovery difficult.
- Form
ChatGPT omitted this.
Loans generally require no special form for effectiveness, but above PLN 1,000 documentary form is required for evidential purposes. A loan with a board member who is sole shareholder requires a notarial deed.
Ancillary security agreements may require particular forms on pain of invalidity.
- Effective repayment security
The template’s statement that the lender receives a third-party surety is insufficient. A written agreement with the surety is needed; if included in the loan contract, the surety must be a party. The analysis neither explains suretyship nor its necessity or alternatives.
Depending on relationships and terms, lending may be conditional on security: surety, ordinary or registered pledge, mortgage, or notarised submission to enforcement. Corporate approvals and other requirements must also be examined.
Where the borrower has marital community property, spousal consent enables enforcement against it. Without consent and separate assets, recovery may be seriously hindered or impossible.
- Conflicts of interest
ChatGPT identified a company risk, but this is a major risk for the board member. Conflicting interests involving the company, member or close persons require disclosure and abstention.
The prevailing view is that breach does not invalidate the agreement but may create the member’s damages liability, particularly on non-market terms.
- Tax obligations
Numerous duties arise, breach potentially constituting fiscal offences. ChatGPT identified none.
The borrower generally pays civil transaction tax and files returns. Interest-free or below-market loans may also generate taxable income equal to saved interest. Exceptions exist: administrative case law may exclude liability in specific shareholder-to-company lending circumstances.
The lender owes tax on interest received.
Complex tax law requires case-specific analysis.
- Additional special obligations
Capital-market rules are one example. Public companies must disclose inside information to protect investors; a financing loan may qualify. ESPI announcements through exchange systems are used. Breach exposes the company to civil and administrative liability and directors to criminal liability.
Other rules govern maximum interest and costs and special contents of consumer loans.
Other issues
Analysis may reveal better legal and tax alternatives. Board members are often shareholders seeking development funding or financial improvement. Commercial law permits structures such as capital increases and debt-to-equity conversion or shareholder supplementary payments, potentially more efficient than a loan.
Confidentiality also matters for trade secrets and sensitive information. Entering data into ChatGPT risks unauthorised disclosure or use. Under its terms, users permit use for providing, developing and improving services[2], effectively allowing broad use. Advocates and legal advisers have professional secrecy and data-protection duties supporting trust and client protection.
Conclusion
The example shows ChatGPT cannot replace lawyers’ knowledge and experience even for small transactions. General information is insufficient against Polish law’s complexity and formal requirements. A generic template may fail in larger or complex deals. Current technology cannot replace professional advice even in a minor, straightforward matter. Receiving answers only to questions asked is problematic where users do not know what to investigate. Before drafting, a lawyer establishes necessary information, including through client interviews.
Case-specific verification makes general advice of limited value. Reliance alone can produce invalidity, damages, administrative, criminal or fiscal liability. Repair may be harder than professional analysis and drafting. Unlike the software provider, a professional adviser takes full responsibility. With ChatGPT, users bear responsibility for reliance and risks associated with input data.
[1] Software at https://chat.openai.com/, developed by OpenAI, L.L.C., automatically generating responses to user prompts from its information.
[2] Terms of use: https://openai.com/policies/terms-of-use.
