Artificial intelligence ("AI") and cloud computing are both popular buzzwords generally referred to as separate concepts without considering the potential for interplay and interconnectedness between the two technologies. Companies could benefit from incorporating cloud computing functionality into their AI operations. For example, AI algorithms can be trained and deployed on a far greater capacity using cloud computing rather than on local servers. Cloud-based AI can also process vast quantities of data through cloud computing infrastructure. This article explores the potential for synergy between cloud computing and AI algorithms while identifying the dissonance that might occur as a result of the collaboration.

The synergy between AI and cloud computing

Cloud-computing and AI tools are individually relied on as technological developments that have provided convenience to users and, to some extent, have disrupted existing industries. When combining the two, the company may experience the following benefits:

  • Scalability: AI algorithms, especially deep learning algorithms, require great computational power. By using cloud-based AI, a company can increase its computational power without any substantial investments in hardware. The cloud-computing component will also allow the company to benefit from upward and downward scaling to suit the company's operational workloads and requirements.
  • Data accessibility: The cloud can provide access to various information sources, which can significantly improve the quality and accuracy of the AI's decision-making ability, given the diversity in data available to the AI solution.
  • Parallel processing: Training, testing and deploying an AI algorithm requires a high degree of computational power. If AI is trained on a local device, it will often only have access to that machine's processing power. However, cloud-based AI is connected to numerous graphic processing units (GPUs) that provide far greater computational power, thereby reducing processing times.
  • Cost effective: As cloud computing can reduce processing times and improve the quality of AI's decision-making, it can lead to a reduction in costs for AI processing.
  • Global reach: Cloud vendors operate on a multinational level, often with data centres spread across the globe. This gives a company access to diverse quantities of data, which can improve the accuracy of the AI algorithms.

The dissonance between AI and the cloud

As with all technological developments and tools, there are various risks associated with the use of cloud-based AI, which includes (without limitation) the following:

  • Functionality and availability dependencies: Linking a company's AI capabilities to the cloud means that AI functionality relies directly on cloud functionality and availability. Cloud downtime can result in performance drops due to latency or, in the worst case, complete shutdown for cloud-based AI deployments. Moreover, transferring data from local servers to the cloud introduces a potential vulnerability that hackers could exploit to access company data. To ensure data security, a company must secure its internal servers and verify that the cloud vendor has robust cybersecurity measures in place.
  • Contracting with cloud vendors: When integrating cloud computing into AI operations, a company must secure a well-negotiated cloud services agreement that safeguards its interests, including security, intellectual property, and data protection. The agreement should also cover service levels and penalties to provide the company with recourse if the cloud vendor falls short. For instance, if the agreed uptime service level is 99.99%, the contract should allow the company to hold the vendor accountable by claiming service credits or penalties for any disruptions or profit loss resulting from a service level default.
  • Cloud security: Due to the inherent risks in moving data to the cloud, a company will need to ensure that it has conducted proper due diligence on potential cloud vendors, their offerings, as well as their technological and security infrastructures. Failure to do so may expose the company to risks of cyberattacks in respect of the data being hosted in the cloud.
  • Contractual bargaining power: A company will need to consider its bargaining power when it comes to negotiating and concluding contracts with cloud vendors. Companies will usually have less negotiation power if it is concluding a contract with a larger cloud vendor. Nevertheless, from a commercial perspective, the company must focus on negotiating the terms and conditions relating to vendor migration options, vendor lock-in, and the costs of switching between cloud vendors.
  • Unauthorised access to personal information: Cloud vendors store large quantities of data, which could contain personal information. A company will need to be cautious when it comes to processing or storing personal information in the cloud and will need to ensure that in doing so, it complies with the Protection of Personal Information, 2013. As part of acquiring cloud-based AI functionality, a company must consider whether it is licenced to process the personal information being made available to it and whether it has obtained the necessary consent to process or store such personal information.
  • Infringement of intellectual property: As with personal information, proprietary material may be stored in the cloud. A company must ensure that it is licenced to process any such proprietary materials. A company should also ensure that its service agreement with a cloud vendor clearly deals with the following issues:
    • provides the company with a license to process the respective proprietary materials;
    • sets out which data the cloud-based AI may access and process; and
    • limits the company's liability to acceptable levels for breaches of intellectual property rights.
  • Unauthorised access: If a company were to connect to a cloud database and begin processing large quantities of data, there is a high likelihood that the company could process or otherwise disclose personal or proprietary information without proper authorisation, which may then result in damages and reputational harm. As it is often challenging to manage visibility over the information processed by AI tools (even more so if it is connected to the cloud), the company must ensure that it has robust contractual agreements in place to regulate its rights and limit its liability. There is also the risk that a company's proprietary, personal, or confidential information could be used as input into the AI tool, potentially resulting in the tool being trained on such information, and using that information as an output going forward. It is difficult to manage this risk, but generally, a company will need to train its staff on the information that is classified as company proprietary, personal and confidential information. The company should seek to "ring-fence" this information in the cloud, in such a way that prevents it from being accessed by or used to train the AI tool.

Cloud-based AI provides an opportunity to significantly enhance a company's AI capabilities and companies that are first to embrace cloud-based AI could obtain a competitive advantage when doing so. However, these advantages may only be realised when a company has carefully considered the cloud marketplace, engaged in vendor due diligence processes, and entered favourable cloud service agreements. failure to maintain proper vigilance over this process could lead to hidden expenses that may outweigh the benefits of cloud-based AI. It is essential to have legal support during this process to ensure that your risk exposure is reduced or mitigated. ENS' Technology, Media and Telecommunications Team has assisted many clients in negotiating and reviewing cloud vendor agreements.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.