Emerging issues Additional issues Open source software (OSS) Given the issues around fairness, transparency and explainability, OSS AI tools have emerged offering lawful basis and data security, not environments for training intelligent to mention the fact that many AI agents using deep reinforcement tools will process special data (eg and imitation learning like the Unity biometric data), game developers Machine Learning Agents Toolkit (ML- and publishers using AI in games Agents). These types of tools have need to consider data protection quickly become popular with game every step of the way. A Data developers. Using open source AI Protection Impact Assessment tools undoubtedly offers benefits, will also likely be required before but also carries risks that need to using AI which processes or roerl addressed to ensure generates personal data. the long-term future of the game relying on them. There are inherent conflicts between using data-driven AI, and See our Open source section for data protection – for example, more information. around data minimisation. The ICO Personal data has already produced extensive guidance on AI and data protection To use AI to improve player with more to come. experience, AI models need to be trained, which can involve tracking See our Data sectionfor more millions of player interactions and information. data points, including those of Ethics and accountability children. This carries inherent risk in an already highly regulated area. AI algorithms are only as good as the data they’re trained on and the The underlying principles which potential for bias in training data, the UK government is proposing which can lead to discrimination or regulators must take into account inequality issues with output, is well when regulating AI overlap heavily documented. Transparency, fairness with the UK GDPR principles. and accountability should govern the 157