Chat GPT

Like all technological changes, the recent explosion in the use of Chat GPT and similar AI models creates new legal and regulatory issues. New Zealand is yet to adopt a specific legal regime to govern the use and control of AIs like Chat GPT, but existing laws will continue to apply. In this article, we explore some factors to consider in embracing the opportunities widely available AI brings.

But first, what is Chat GPT?
Chat GPT is a large language model. This is code that allows a computer to understand patterns in language. By analysing 175 billion examples of human communication from the internet, textbooks and elsewhere, it can predict the likelihood of the ‘correct’ next word in a sentence.

Doing this, Chat GPT can answer questions from users in a conversational manner. It can also produce text in any conceivable format on request, within seconds.

For example, it can write articles, draft a business plan, write and debug computer code, draft an article or essay, provide a recipe for a cake or cocktail, and even attempt poetry and song lyrics.

These uses have caused widespread speculation about the possible uses of AI to replace internet search engines and many reference texts, and that it may threaten content creation type roles.

A number of major companies like Alphabet (Google), Meta (Facebook), and Alibaba have announced competing models will be released in the near future. Some of these are based on alternative AI technologies that will increase the power and possible uses for large language models. What we say here about Chat GPT will equally apply to any of these other engines.

How accurate are these large language models?
Like all computer systems, Chat GPT’s output is only as good as its input. As computer engineers have long said “garbage in, garbage out”. While having 175 billion examples to draw on allows Chat GPT to be impressively accurate in dealing with familiar topics, it is limited to drawing on that data set.

Much of this comes out of the United States, reflecting what is available online. This means that Chat GPT has limits answering questions dealing with New Zealand specific conditions and situations. Also, Chat GPT is less accurate providing information about specialist and technical topics not comprehensively discussed in publicly available sources.

Chat GPT (and all similar models) carry clear warnings about not relying on its output as a substitute for professional or expert advice. Open AI (Chat GPT’s creator) accepts no liability for harm caused by relying on Chat GPT’s answers.

Users should be careful to verify the information from Chat GPT, and seek relevant advice, before relying on it. This will be especially true of information about living people and current events.

If the user is intending to pass on the information from Chat GPT to others, they may find themselves held liable for any inaccuracies in the output.

Who ‘owns’ the input and output from Chat GPT and similar models?
AI language models like Chat GPT are not people and do not “own” the content they generate. Who owns that content will depend on the licence user agreement for the particular model being used.

In the case of Chat GPT, Open AI (the creator) assigns its intellectual property rights in the output to the user. This means users are not at risk from Open AI claiming copyright infringement from the use of Chat GPT output. That may differ with other models, and it will be necessary to review the terms and conditions in each case.

However, Chat GPT does draw on copyright material from others and may include this in its answers. Users should check their output is not plagiarised before using it, especially in any commercial contexts or where using that material for profit.

Confidentiality and privacy
Users of Chat GPT and similar models need to be aware their inputs become part of the model’s data set, and can appear in outputs provided to other users.

This is particularly important with commercially sensitive information. Samsung workers in Korea looking to have Chat GPT debug their computer code inadvertently cost their employer trade secrets protection when the code they submitted later appeared in responses to other users.

Similar concerns exist around disclosing personal details and other private information to Chat GPT. The Canadian Privacy Commissioner has launched an investigation into Chat GPT concerns. In New Zealand, users and businesses’ obligations under the Privacy Act 2020 around the release of information to third parties will apply when using Chat GPT.  Information that could identify individuals should only be released for purposes reasonably necessary for the user’s business. It must also fall within the specific uses for the information identified at the time the information was collected.

Being honest about using AI at work and school
Even if copyright, confidentiality and privacy issues do not arise, people should be cautious using Chat GPT and other models and presenting it as their “own work”.

There has been particular concerns from universities and schools about students using AI models like Chat GPT to produce assignment work. Some systems, like the International Baccalaureate, have said use of Chat GPT is acceptable so long as it is cited correctly.

Meanwhile, the Queensland government has banned it from schools there. New Zealand schools and universities are each developing their own response. Students will need to follow their own institution’s guidelines.

In the office, a range of other legal issues can arise from employees using Chat GPT without telling their employer. Surveys indicate large numbers of employees are already using Chat GPT surreptitiously. Employers and employees are required to deal with each other in good faith. Employees could be seen to be acting deceptively in using Chat GPT to substitute for their own work: especially if in a role where their own creativity or judgement is a key part of their function.

Employers should turn their mind to creating formal policies about the use of Chat GPT by staff to make the most of new opportunities while setting clear boundaries.
Since 1937, Holland Beckett Law has been helping New Zealanders navigate the ever-changing world with timely and practical legal advice. Our dispute resolution and employment law teams can help you understand how the existing law will map onto tomorrow’s world.