What PocketAI Customers Should Know About the EU AI Act

The European Union has come to a consensus on a legislation they have been diligently working on for several years concerning the advancement and utilization of AI within the EU. The final enactment will not come into full effect until at least 2025.

The legislation adopts a risk-based approach, which means it evaluates the level of risk associated with different types of AI and establishes regulations for each category based on that risk assessment. Technologies such as facial recognition have been identified as high-risk. On the other hand, large language models have been classified under the limited risk category.

According to this law, companies and organizations that develop large language models would be required to comply with transparency guidelines regarding their foundational models. This includes creating technical documentation, adhering to EU copyright law, and providing detailed summaries of the content used for training purposes.

When it comes to applications and users utilizing these underlying models, there are two important points to consider:

• Firstly, providers must guarantee that AI systems, such as AI chatbots usi, which are designed to interact with human beings, are developed and designed in a way that makes it clear to individuals that they are engaging with an AI system, unless this fact is already obvious from the circumstances and context in which the system is being used.
• Secondly, users of an AI system that creates or alters visual, auditory, or video content which significantly resembles existing individuals, objects, places, or other entities or events such as DALL-E 3, and could deceive someone into believing it is authentic or truthful (commonly known as a 'deep fake'), must disclose that the content has been artificially generated or manipulated.

It is important to note that these regulations will not be implemented until the year 2025.

What implications does this have for the average user of PocketAI?

The majority of the EU AI Act primarily focuses on regulating systems that are considered high-risk to the individuals within the EU, such as facial recognition technology. However, large language models like the ones utilized by PocketAI are categorized as "limited-risk" and are subject to less stringent regulations. Instead, they must adhere to transparency requirements and comply with EU copyright law before releasing a new foundational model.

So, what does this actually mean for the average user of PocketAI? Well, when it comes to end-users of AI, the Act proposes minimal regulation. There are two noteworthy points to consider.

• Firstly, it is explicitly prohibited to employ AI for the purpose of creating deep fakes or falsely impersonating someone. This prohibition is not new, as it already aligns with PocketAI's terms of service.
• Secondly, if you are a PocketAI Enterprise user and utilize PocketAI's API to power a chatbot or a similar application, it will be mandatory by 2025 to provide clear disclosure within the chatbot experience. This disclosure will inform users that they are engaging with a chatbot rather than a human.

It is important to note that these are our initial interpretations of a rule that will be further elaborated upon in the upcoming year. It is essential to understand that this information does not constitute legal advice. For complete clarity and guidance, it is always recommended to consult with your own legal team in addition to this information.