The Use of Artificial Intelligence (AI)

Artificial intelligence (AI) is the simulation of human intelligence processes by machines, most prominently, computer systems. AI works by identifying patterns and connections within large datasets. The most relevant AI products for our work and everyday life are chatbots, often framed as “assistants” such as OpenAI’s ChatGPT, Google’s Gemini, and Microsoft’s Copilot. When we use the word AI in this policy and our work generally, we predominantly mean AI chatbots.  

These programs imitate human conversation with the user. They predominantly use algorithms trained on large datasets to match a user’s prompt to a specific “intent” and respond accordingly. 

AI can be a valuable tool that supports our work and is best used as that, a tool and not as a substitute for genuine ideation, research, etc.

However, the “AI supply-chain” also plays a significant role in damaging the world’s ecosystems and will, in the future, be a catalyst for large-scale climate displacement and migration. Furthermore, they are built on the appropriation of others’ hard work, with copyright laws disregarded, with few compensated.  

We are also aware of the troubling trends of AI facial recognition tools in the asylum system, although not directly related to our remit, we do not want to contribute to the normalisation of tools that cause so much harm in our field and perpetuate racist and colonial narratives. 

The Law Society’s list of risks related to generative AI provides a relevant summary of several of the potential negative consequences of integrating AI chatbots into our work. This list is not exhaustive. 

• Intellectual property risks: potential infringements of copyright, trademarks, patents and related rights, and misuse or disclosure of confidential information
• data protection and privacy risks: concerns related to the unauthorised access, sharing or misuse of personal and sensitive data
• cyber security risks: vulnerabilities to hacking, data breaches, corruption of data sources and other malicious cyber activities
• training data concerns: the use or misuse of data to train generative AI models, which could result in biases or inappropriate outputs
• output integrity: the potential for generative AI to produce misleading, inaccurate or false outputs that can be misconstrued or misapplied
• ethical and bias concerns: the possibility of AI models reflecting or amplifying societal biases present in their training data, leading to unfair or discriminatory results. There may also be environmental, social and governance (ESG) considerations.
• human resources and reputation risks: if the use of generative AI may result in negative consequences for clients, there may be reputational and brand damage 

The lists below are by no means exhaustive, rather they address areas relevant enough to our day-to-day work to have their use or lack thereof specified.

AI tools can be effective at improving some processes, making them more efficient and accurate. Such tools should be used purely as a time-saving mechanism and should not be a substitute for seeking feedback from a colleague.

What we can use AI for  

• Administrative tasks like auditing spreadsheets, organising folders, etc (with no personal information)  
• Requesting feedback for written and visual work but does provide substantial edits that you can plug-in (with personal information removed)
• Early brainstorming/stress-testing of ideas
• Suggesting alternative language/wording, for example, suggestions to make a sentence more informal or formal
• Identify grammatical errors in our writing (personal information removed) 

The House of Commons suggested AI Workflow can be used as useful scaffolding when undertaking the tasks above:

1. Define the task clearly and decide whether AI is appropriate
2. Use AI to do the task
3. Check all factual claims against reputable, authoritative sources
4. Review outputs with someone who has subject expertise
5. Edit, contextualise and take responsibility for the final content

What we will not use AI for  

• Research in pursuit of definitive factual answers
• To generate images or audio
• Interpreting case law or policy
• Any process that includes personal information
• Any process related to assisting those looking for support
• Translating documents in cases where interpreters are available, recommended or required
• Final copies of any piece of work (social media posts, final drafts of research, web posts, etc)  

Where possible, limit your use of AI tools; most public AI tools are imprecise and prone to hallucination. This is most important during the research period, when they often create fake citations, fake research, fake case law and are ultimately unable to capture fast-pacing domains like immigration law/policy.   

Under no circumstances should AI be used in any processes that may disclose personal information.

Chatbots remember every prompt that you provide and use this information to inform other queries; assume that anything you type could become public knowledge. 

It is important that we do all that we can to reduce the “AI-ness” of our published work. Many who may engage with our work can and will be put off by work that appears AI-generated even if it has been written by a human.

The following is a list of indicators that a piece of work may have been AI-generated; none of these, in isolation, should be avoided. Despite that, please be vigilant in minimising use, especially in combination.  

• Em dashes (—)
• The use of the following sentence structures
  • • “It’s not just about X, it’s about Y”  
    • “This is more than a XYZ” 
    • “Taken together”  
• The overuse of semicolons, often in place of conjunctions
  • • “Social media fosters global connectivity; it also serves as a vital tool for sharing work”   
•  The use of emojis and bullet points in an excessive way – often together
  • • 🧠 Understanding AI: What it is and isn’t. 
    • ⚡️ AI’s Impact: Efficiency gains.
    • ⚖️ Ethics: The need for responsible us  
• The excessive use of the rule of three – for example  
  • • “Blood, sweat, and tears”  
    • “The party of the people, by the people, for the people”
• Overly polished and monotonous tone of voice
• Overly confident prose that lacks evidence or explanation 
• American spelling (“color”, “emphasize”) in documents where British English is expected.