A customer-facing AI assistant tells a customer that a refund is available under a policy that does not exist. The customer relies on it. When the company refuses, the customer complains, and the board hears about it from a newspaper. In the first meeting afterward, three groups each assume someone else owns the problem: the business unit, IT, and the vendor.
Directors need to settle this question before an AI app fails, not after. Choosing the right AI app development company is also important, but accountability cannot rest with the vendor alone. This guide explains where accountability tends to land, why the vendor is not a shield, how to map responsibility inside your company, what to put in contracts, and which questions to ask. It is general information, not legal advice, and rules differ by country. It is written for directors and senior leaders, with no technical background required.
The Short Answer: The Deploying Organization Is Usually Accountable
In the widely cited case, a Canadian tribunal in Moffatt v Air Canada, 2024 BCCRT 149, rejected the argument that the airline’s chatbot was a separate entity and held the company responsible for all information on its website. The tribunal ordered the airline to pay C$812.02.
The amount was small, but the principle matters. Under current law in the US, UK, EU, and Canada, AI systems have no legal personality and cannot themselves be held liable. The ruling is a useful example, not a universal rule, and outcomes depend on jurisdiction and facts. Still, it signals how regulators and courts tend to think: if your organization puts the AI in front of customers, expect to answer for what it says.
The Vendor Is Not Off the Hook, but Not a Shield Either
Using a vendor does not move your accountability to the vendor. The vendor may share responsibility, but your customers deal with you. The revised EU Product Liability Directive treats software, including AI, as a product, and increases the vendor’s exposure, but as an additional defendant, not a substitute one.
The directive applies to products placed on the market from 9 December 2026, covers all types of software including AI systems, and allows manufacturers to be held liable for a defect that existed at release, with new tools for courts to request evidence. Deployers also carry duties under the EU AI Act. If your company builds, customizes, or substantially modifies an AI system, ask counsel whether you could count as a manufacturer yourself.
Map Accountability Inside Your Company
Law aside, the board needs to know who in the organization answers for each AI system. Ask:
- Who is the named executive owner of this system?
- Who approves launch, major changes, and the use of new data?
- Who decides when to pause or switch it off?
- How do problems reach the board, and how fast?
A good answer names a person, not a department. Define decision rights in writing: the business owner is accountable for outcomes, technology for operation and security, legal and compliance for rules, and the vendor for what the contract specifies. Boards should also know where AI oversight sits, whether with the full board or a committee, and what information they will receive. Vague shared ownership is the most common reason failures linger.
Contract Questions: Who Pays, Who Fixes, Who Proves
The contract is where accountability becomes concrete. Ask legal counsel to review these points before signing:
- Responsibility for defects. What is the vendor responsible for if the system is faulty, and how are remedies defined?
- Liability and indemnities. What limits and protections apply, and do they cover regulatory, data, and third-party claims?
- Service levels. What response and fix times apply to serious errors?
- Incident duties. How quickly must the vendor tell you about problems, data exposures, or model changes?
- Audit and records. Can you see logs, test results, and version history?
- Ownership and exit. Who owns code, models, and data, and what happens if you leave?
Also review insurance coverage, since some policies may not clearly address AI-related losses. Contracts allocate risk between you and the vendor, but they do not bind customers or regulators.
Evidence and Incident Readiness
When something goes wrong, the quality of your records shapes the outcome. Keep documentation of testing before launch, approved changes, the model and prompt versions in use, logs of important decisions, and complaints and incidents. Balance this with privacy rules on data retention.
Prepare an incident plan before you need one. One legal advisory warns that traditional cyber incident playbooks may be insufficient, because AI incidents can involve model behavior, data leakage through prompts and logs, third-party model providers, and rapidly changing outputs. Your plan should say who investigates, who can disable the system, how customers are told, and who speaks for the company. Rehearse it once. Providers also retire older models, and Anthropic says it gives at least 60 days of notice before retiring publicly released models, so assign someone to track provider notices.
A Director’s Accountability Checklist
Before approving, or reviewing, any customer-facing or high-stakes AI app, ask:
- Is there a named executive owner, and does the board know who it is?
- Is the system classified by risk, and are the rules in each market identified?
- Are testing, version, and decision records being kept?
- Does the vendor contract cover defects, incidents, audit access, and exit?
- Is there a tested plan to pause the system and communicate with customers?
- Has legal counsel reviewed liability, insurance, and disclosure duties?
If the answer to any is “not sure,” treat that as the agenda for the next meeting.
Why Choose Quytech
Quytech describes itself as a technology partner specializing in AI development, generative AI solutions, and mobile app development. Its Clutch profile also lists agentic AI solutions, data engineering and AI analytics, and IT staff augmentation among its services.
Directors can put the checklist above to Quytech, as they would to any vendor. Ask what it is responsible for under the contract, what logs and test records you will receive, how incidents are reported, and what support and exit terms apply. A suitable partner should help you clarify requirements, plan development in stages, and answer these questions plainly.
Frequently Asked Questions
Q1. Can we blame the AI vendor when the app fails?
You may be able to share responsibility with the vendor through the contract, but customers and regulators generally deal with the organization that deployed the app. Treat vendor terms as a way to allocate risk between you, not to remove your accountability.
Q2. Does the EU Product Liability Directive apply to us?
It applies to products placed on the market from 9 December 2026 and mainly concerns software manufacturers, including AI. If you build or substantially modify AI, you could be treated as one. The rules focus mainly on harm to individuals, so ask counsel how they apply to you.
Q3. What records should we keep?
Keep pre-launch test results, approved changes, model and prompt versions, important decision logs, complaints, and incident reports. Set retention periods that respect privacy law, and make sure you have access to the vendor’s logs under the contract.
Q4. Do disclaimers protect us if the AI is wrong?
Disclaimers support transparency but do not replace accuracy. In the Air Canada case, the airline’s attempt to distance itself from its chatbot failed. Seek legal advice on wording, and focus on testing, review, and escalation paths.
Q5. Who should own AI risk at the board level?
It varies. Some boards keep oversight with the full board; others use a committee. What matters is clarity: a named executive, regular reporting, defined decision rights, and a way for problems to reach directors quickly.
Conclusion
When an AI app fails, accountability usually lands on the organization that deployed it. Vendors can share responsibility, and new rules such as the revised EU Product Liability Directive increase vendor exposure, but they add defendants, not remove your own.
Directors can prepare by naming owners, writing decision rights, tightening contracts, keeping evidence, and rehearsing incident response. If you want to discuss your project with an experienced technology partner, Quytech is one option worth including in your comparison. Settle accountability before the failure, not during it.