Singapore: AI ethics guidelines for lawyers

[SINGAPORE] The Ministry of Justice (MinLaw) said in new guidance published on Friday 6 March that lawyers who use generative artificial intelligence (AI) in their work will continue to be ultimately responsible for all work produced as part of their professional duties to clients.

While AI offers powerful capabilities to aid legal work, the technology has inherent limitations, and professional responsibility remains with lawyers, who must use their expertise to guide and validate the results generated by AI, the guide said.

“Use of the Gen AI Tools does not delegate or relieve you of these obligations.”

The Guide to the Use of Generative AI in the Legal Profession, announced at the MinLaw event at Raffles City Convention Center on Friday, is the first framework setting out the ethical use of AI in the legal profession. The scheme is open to anyone practicing law in Singapore, including private practice lawyers, in-house lawyers, paralegals and law students.

Although not codified, the guide sets out three fundamental principles that legal professionals should keep in mind when using AI in their work: professional ethics, confidentiality, and transparency.

It also lays out a five-step implementation framework for law practices and legal teams, from developing AI governance policies and assessing workflow needs to evaluating tools and training staff.

Justice Minister Edwin Tong called AI “the most disruptive force for the legal profession” and said the government would “fully support” companies adopting the technology.

“We stand ready to invest as you make change, and we intend to put meaningful support on the table to help the legal industry make this change,” he said.

In devising the guide, the government decided not to hard-code it at this time to avoid stifling innovation, Tong said. “I don’t think now is the right time, because in terms of technological evolution, we are still far from where we are.”

Targeted support will also be provided to small and medium-sized enterprises facing the challenges of economies of scale in implementing AI implementation.

The Department’s Legal Innovation and Future-Ready Transformation initiative will advance these efforts by helping businesses analyze their needs and understand which products are best suited for them.

“I don’t believe that AI will replace human lawyers, at least not in the foreseeable future, but those who adopt it and make better use of it will replace those who don’t,” said Tong.

hallucinations and prejudices

The main concern this guide addresses is AI hallucinations, or the risk that the output is inaccurate or fictitious.

While you can’t completely eliminate such errors, you can reduce their chances by, for example, feeding the AI ​​with reference documents specific to the AI ​​to anchor its responses, rather than drawing freely from training data.

The guide also warns of bias risks due to training data that can reflect historical bias, unrepresentative samples, and algorithm design choices that can distort legal inferences.

Legal professionals are therefore encouraged to test AI output across different case types and client groups and ask AI to explain the reasoning behind its recommendations.

The guide also recommends aligning the level of human oversight with relevant stakeholders.

For example, for high-risk work such as court filings or legal advice, lawyers will need to review and approve AI-generated work before it can be used.

For more routine tasks, such as client updates or meeting notes, a sampling check may be sufficient.

Confidentiality and transparency

Lawyers’ duty to protect client information also extends to artificial intelligence tools, the guide says. The report recommended that companies prioritize enterprise-grade tools over free public platforms for sensitive data and secure commitments with vendors that prohibit the use of customer data in training AI models.

When using free tools, attorneys should anonymize data and double-check that data retention settings are disabled.

Regarding transparency, lawyers must disclose the use of Gen AI to their clients if the technology is materially used to create a work product. You should also do so if it affects the cost of legal services or if the tool’s data processing practices may conflict with your client’s preferences.

Clients should also be provided with the option to opt out of the use of AI.

Implementation of AI

The guide also provides step-by-step recommendations for implementing the use of AI.

Companies must first establish clear governance policies. This means defining the tools that are approved for use, the types of data that can be input into them, and who is responsible for overseeing them.

Companies should also have protocols in place to communicate their AI practices to customers and procedures for reporting errors and data breaches.

Next, companies should evaluate where Gen AI can add the most value to their existing workflows, weighing the risks and feasibility of each use case.

The third step is choosing the right tools. The guide recommends that companies conduct thorough due diligence on vendors, examining their data security measures, how they handle customer data, and whether the accuracy and reliability of their tools have been tested in a legal context.

Companies should start with basic AI tools like Microsoft Co-pilot and LawNet AI before expanding to off-the-shelf legal AI products. Cutting-edge companies may eventually develop in-house AI solutions customized to their specific needs.

Once a tool is selected, companies must implement it in stages. You should start with a pilot group to gather feedback and adjust prompts and workflows before implementing it more broadly.

Finally, companies should regularly review whether Gen AI tools still meet their needs, stay informed of new developments, and update internal policies accordingly.