Dr. Richard Browne:

Fritz Lang's "Metropolis" was made in 1927 and it is essentially about AI. The idea of thinking machines is not new but there are implications for society from the latest developments. ChatGPT is a very public example but there are much more interesting and much larger issues happening in the background, some of which will become public in the next three or four months. To begin with and to answer the question directly, we cannot ban anything. We do not have the power to prohibit anything. We provide guidance and support, which is what we have done in respect of social media, which was raised in previous questions.

In this case, and this has crept into the media to some extent, it is not true that the use of ChatGPT by a Department poses a cybersecurity risk to that Department. The primary issue is whether the data that is being applied to it or any data being released to it was sensitive in any way, shape or form. That is already Departments’ data protection policies. From a cybersecurity perspective, there is no dramatic risk from the use ChatGPT right now. We published a blog on this recently that pointed out that threat actors, as distinct from civil servants, using ChatGPT or other generative AI models could pose a risk. That is very different from the Civil Service using it and posing a risk. From our perspective, the real risks in this are because threat actors, criminals and states can become much more productive and effective in developing their tool sets, be that writing code, writing malware or generating much more believable and coherent narratives for phishing. Those are the real risks from ChatGPT. Individual Departments and agencies bear their own risk; they write their own IT policies. We will publish a piece of guidance on the use of generative AI models in the next little while. It is also important to point out that much of the challenges here are for other entities. Data protection is for individual Departments and the Data Protection Commission, DPC. The DPC is involved in a European-wide project right now looking at the risks of these kinds of issues. There are other questions, such as the types of intellectual property used to inform and build these models.

Like I said, this is a whole-of-government challenge. AI is explicitly referenced in our 2019 strategy. The Government has had an AI strategy for three years at this point. This is not the first time Government has thought about these issues. This one has happened to all of us suddenly, but we have been working on it for a long time.