Contracts & AI
As of 2026, many Ontario business owners have gotten into the practice of drafting their legal documents with an AI tool. Privacy policies, website terms of service, contractor and employment agreements, non-disclosure agreements, customer contracts and shareholder agreements are all tempting legal documents to ask AI to draft for you.
Sooner or later, a deliverable is missed, a payment is disputed or a relationship ends badly, and the gaps AI left in your contract start to cost you. That is when you find out whether drafting with AI was worth the risk.
A contract becomes binding in Ontario when the ordinary elements of formation are present:
A contract drafted by a lawyer, copied from a template site, scrawled on a napkin or generated by a large language model is assessed on the same criteria. The question is not whether AI is allowed to write your contract. It is whether what AI wrote actually does the job.
AI is very good at hallucinating. An AI tool may invent a legal doctrine, cite legislation that does not exist, refer to the wrong jurisdiction, or generate a clause based on a misunderstanding of the law, all while presenting the information with complete confidence.
This is particularly dangerous in contracts because the error may be buried inside otherwise professional-looking legal language and go unnoticed until someone tries to rely on the provision. In some cases, an AI-generated clause may be legally unenforceable, leaving a business without the contractual protections it thought it had. Depending on the nature of the error, it could also create ambiguity or uncertainty that leads to costly disputes over the interpretation or enforceability of the agreement itself.
Privacy policies are perhaps the clearest example of why generating legal language is not the same thing as doing legal work. You can tell an AI tool: “Draft a privacy policy for my Ontario e-commerce business.” It can probably produce one in seconds. But a compliant privacy policy has to reflect the business’s actual data practices.
Before a policy can be written, someone has to answer questions like these:
Under PIPEDA, organizations are responsible for personal information under their control and must address obligations including accountability, identifying the purposes for collection, consent, safeguards and openness. The Office of the Privacy Commissioner of Canada also states that organizations generally need meaningful consent for the collection, use and disclosure of personal information, meaning people must be able to understand the nature, purpose and consequences of what they are agreeing to.
A generic privacy policy containing impressive statements about “industry-standard safeguards” or categories of information that your company does not actually collect can be worse than merely generic. Your published policy may describe a privacy program that does not match the one your business actually operates.
Privacy non-compliance can expose businesses to regulatory investigations, legal disputes and potentially significant financial penalties. Under PIPEDA, certain offences, including knowingly failing to comply with mandatory data breach reporting requirements, can result in fines of up to $100,000. Businesses that use customer information for email marketing must also consider Canada’s Anti-Spam Legislation (CASL), which provides for penalties of up to $10 million per violation for businesses.
The problem is that AI cannot independently verify whether your business’s actual practices comply with these laws. A privacy policy that looks legally sound on paper will do little to protect your business if the practices behind it are not compliant.
Employment and contractor agreements are particularly risky to generate with AI because employment laws vary significantly between jurisdictions. For example, Ontario prohibited most employee non-competes through Part XV.1 of the Employment Standards Act, 2000, effective October 25, 2021, subject to limited exceptions. Yet an AI tool may still generate a polished non-compete clause because these provisions remain common in other jurisdictions.
Even seemingly minor drafting errors can have serious financial consequences. In Waksdale v. Swegon North America Inc., 2020 ONCA 391, the Ontario Court of Appeal held that an unlawful termination-for-cause provision rendered the agreement’s termination provisions unenforceable, even though the separate without-cause provision was otherwise lawful. For employers, this means the contract no longer limits what a terminated employee is owed. The employee is instead entitled to common law reasonable notice, which is often many months more than the Employment Standards Act, 2000 minimums the contract was meant to cap them at.
The same risks extend beyond termination clauses. AI may inadequately address intellectual property or confidentiality, or label someone an “independent contractor” when the reality of the working relationship suggests otherwise. A contract can sound legally sophisticated without actually reflecting Ontario law or the relationship it is supposed to govern.
The financial consequences can be substantial. An unenforceable termination clause may leave an employer liable for months of additional compensation that it never anticipated paying. Separately, businesses that actually violate Ontario’s Employment Standards Act, 2000 can face regulatory penalties, including fines of up to $100,000 for a corporation’s first conviction, rising for subsequent convictions to a maximum of $500,000. For small businesses, these are risks that can far outweigh the cost of having an employment agreement properly drafted or reviewed in the first place.
A shareholder agreement can look deceptively easy to generate with AI. A tool can quickly produce familiar provisions dealing with voting, management, share transfers, confidentiality and dispute resolution. But the real value of a shareholder agreement is not having the right headings. It is anticipating what could happen between the specific people who own the business. A properly tailored agreement should address questions like:
The answers depend on the company’s ownership structure, the founders’ contributions and responsibilities, financing plans, and what the shareholders actually want. AI can generate clauses addressing these subjects, but it cannot decide what arrangements make sense for the particular business and its owners.
The choice does not have to be between ignoring AI altogether and relying on it to do the work of a lawyer. For business owners, AI can be a useful tool for getting organized before a meeting with a lawyer, learning unfamiliar legal terminology and arriving at a legal conversation with better questions and a clearer sense of what you need.
For lawyers, AI can be very helpful in making legal services more efficient, accurate and accessible. What it cannot replace is the judgment behind a good contract: understanding how a business operates, identifying the risks that matter, applying the right law and making deliberate drafting decisions. That judgment, and the responsibility for it, still sits with the lawyer.
At Align Counsel, we work with founders and growing Ontario businesses to make legal support practical, understandable and proportionate to the actual risk involved.
We are a client-focused firm built to move at the pace of the quickly scaling businesses we serve. We believe technology should enhance the quality of legal services and make them more accessible.
Whether you are putting an agreement in place for the first time or revisiting one you already rely on, we can make sure it reflects your business and Ontario law. The first conversation is on us.
Book a free consultationThe information above is general in nature and is not legal advice. Every situation and transaction is different, and advice tailored to your specific circumstances is required to address your particular needs. If you have questions, contact Align Counsel at [email protected].