MP – Heritage organisations are trust- ed knowledge institutions. As such, this means that key values for the sector will involve presenting accurate information to the public, critical reflection, clear provenance information and compliance with the law – for example.
Working with AI can represent a chal- lenge because it is not ‘industry standard’ for AI developers to disclose the source of the data they use to create their AI tool, and the regulatory framework of AI is unclear. So, to mitigate reputation or financial risks, heritage organisations may need to be very selective when choosing AI tools or partners providing them with AI technologies, in the same way that most heritage organisations perform due diligence before acquiring a new item in their collection or accepting sponsorship.
In the briefing you wrote: “Heritage organisations with experience using AI stress the need to improve on core digital infrastructure like data management and storage as well as workforce development and volun- teer training to use the technology effectively and safely.” Which of these do you think will be most problematic?
MP – In my opinion, workforce develop- ment is where it starts, so data infrastructure, storage and digitisation activities (all of which is needed before engaging with AI meaningfully) can happen in a sustainable way.
You list the risks: bias, discrimi- nation, misinformation and rights infringement. Which of these pose the biggest or most expensive threats to organisations?
MP – There is not a single answer for this question, as it will very much depend on what content AI is used on, or to perform what tasks. Not all risks will always fea- ture or appear to the same degree when you use AI for different purposes. If you use AI internally only (to analyse key per- formance metrics for example, like visitor data, or other) inaccuracy would be your main risk. If you use AI to detect objects or text within collections then bias would likely be the name risk, together with inaccuracy. If you present this information to the public, then bias and misinformation would be your main risks because content could be used out of context for example. And so forth.
How significant is rights infringe- ment and what forms can it take – were you surprised by any?
MP – Rights infringement is very signifi- cant because it is likely that most AI tech-
34 INFORMATION PROFESSIONAL
nology has been developed by mining data online, and it is not clear whether that is lawful or not. So, rights-infringement could have been ‘baked-in’ to the technol- ogy itself, and that’s even before you start using (or misusing) the tool later on. A common type of infringement is uploading information which is either personal or confidential data onto AI tools like ChatGPT, without appropriate consent, in order to produce summaries, reports or write emails. This is not an appropriate, or recommended use, of this type of content, unless you have appropriate safeguards to maintain the confidentiality of the information and prevent further processing. It is not likely that off-the- shelf tools like ChatGPT provide these safeguards.
Is it right to say the legal/regulatory guidance is still too vague? If so, what sorts of extra risks does this pose?
MP – At the moment there is little to no guidance relevant to the heritage sector on the lawful or safe use of AI. This is partly because this guidance will need to come from the courts, and this takes time. And it also because the UK government has opted for a ‘pro-innovation’ strategy for AI, which involves a low-to-no-regula- tion approach to the technology. Having said that, guidance is underway from other corners. For example, the EU AI Act has recently been passed and may set new industry standards within EU-members which may influence prac- tice in the UK, including for the heritage sector. Similarly, individual governmental bodies in the UK are likely to produce guidance to support organisations with specific issues. For example, ICO in the UK is doing excellent work in supporting and educating organisations across all sec- tors on AI and personal data protection.
Mathilde Pavis. Photo ©Melanie Bir
Are there actions the sector can taketo minimise the risks caused by unclear legal and regulatory advice?
MP – Understanding AI and working with AI (should an institution choose to do so) will be a learning curve for any- one, whether you work in IT, marketing, or collections management, whether you are professional, a hobbyist or volunteer, new to the heritage sector or nearer the end of your career. With this in mind, it would be immensely helpful for the heritage sector to pool resources to shorten each other’s learning curve, and this starts by sharing experiences. What works and what did not work? And having the networks and forums to do so. Then comes investments in digitisation tools and data management services which could also be mutualised to lower the costs, and then comes the meaningful deployment of AI within organisations.
Are you working on anything else now that might be of interest to pro- fessionals dealing with copyright?
MP – At the moment, I either work one- to-one with organisations who want to prepare for AI or have developed an AI tool and want to make sure it is used legally and ethically. And I’ve now taken this learning and turned it into valuable lessons learnt and into online courses for people wanting learn in their own time and at their own pace. The courses are not out yet, so watch this space! People can find out more about my work on my website
www.mathildepavis.com or get in touch directly. IP
l Book your place for CILIP’s Copyright Conference 2024 at
www.cilip.org.uk/ CopyrightConf24
April-May 2024
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56
