AI and Law - AI Tech Transaction Agreement (part 2) (ENG)
Limitation of Liability
Almost all license agreements include the "Limitation of Liability" clause. There are usually two steps of limiting damages. The first is limiting the liability for indirect, consequential, and incidental damages (i.e. type of damages) and the second is limiting the liability to a certain amount (i.e. liability cap) even if direct and general damages are occurred. It usually multiplies the fee paid by Licensee. For Licensor, it is advantageous to define both steps of liability, while for Licensee, it is advantageous to carefully consider the second step of liability even if they acknowledge the first one. This is because, if the Licensee ultimately provides the service to consumers, the Licensee may be required to assume unlimited liability to consumers of that service (especially for general individual consumers).
However, if the Licensee request that the liability cap should be lifted altogether, the negotiations are likely to be stalled. Even if there are restrictions on the types of damages in the agreement, it is very difficult to clearly distinguish between direct damages and other special damages (such as indirect, consequential, incidental damages, etc.) unless a lawsuit is filed, and the above restrictions in the agreement often do not function properly in practice. As a result, in reality, lawyers often include all the ambiguities of the damages in the agreement at first and engage in intense negotiations between the parties with a final share (ex. 60 percent to 40 percent) after. The presence of restriction on the amount is a crucial bargaining point for the Licensor, so it is highly likely that they will not yield this clause, and deciding the specific amount of liability cap will be the main battling point.
The process will be more intense especially for AI technology licenses. In AI technology transactions, failures or malfunctions in hardware or systems can cause very serious consequences. For example, if a system that collects and analyzes highly sensitive personal information, such as one's medical history, is down and the information is exposed unencrypted to other users, the amount of damages required as a result of class action suits can be enormous and it will be difficult for Licensors to expect to bear these risks indefinitely.
Even in general technology license agreements, it often requires either or all of the parties to purchase appropriate insurance. For instance, a performance bond insurance serves to protect the other party if either party fails to perform its contractual obligations (including liability for damages). Of course, the types of insurance required will vary depending on the needs and types of contracts for each party. Typical insurance includes the comprehensive general liability, employer's liability, and workers' compensation, in addition to performance bond insurance or maintenance bond insurance.
The importance of insurance increases in areas that are just being explored, such as AI technology. When damage or loss occurs due to an event, revealing the causal link between the event and the loss is the basis of the compensation but people still lack experience in identifying the causal link. Until sufficient experience is built up, the causal link between most events and losses will often be concluded unclearly, and as a result, people will rely more on insurance.
However, there are no specific data or research results on how the previously mentioned AI system can cause casualties or property damage to third parties. Insurers will also need empirical research on which damages can be caused during the data collection, accumulation (storage), and analysis process in which AI systems operate. They need to classify errors, 1) in the collection stage, 2) in the storage stage, 3) in the analysis stage, and therefore, more and more data experts are getting employed for the job.
For the parties signing the agreement, it will be necessary to examine whether the insurance product to which they purchase can cover the damage caused by the AI system (or if it should be recognized as force majeure) and whether the coverage amount is sufficient to cover the loss or damages. In addition, it is also recommended to check if the other party’s insurance protects me appropriately in case the other party fails to perform or pay, and whether receiving double protection from more than one insurance is necessary.