Do employers need a ChatGPT policy?

“I think employers need to get out ahead of it.”

So says Ritu Mahil, partner at Lawson Lundell in Vancouver, in talking about the latest issue confronting HR: the rise of open-source AI such as ChatGPT.

The tool is sweeping the nation, as they say — and implications for the workplace are only starting to be understood.

One recent survey found that 49 per cent of employers are using the tech innovation — for tasks such as job descriptions, writing code and research — while another 30 per cent plan to do so.

On the other hand, several employers are clamping down on the use of ChatGPT among employees, such as Amazon, Verizon, Citigroup, Goldman Sachs, Wells Fargo and Accenture.

But does this make sense? Or is it better to craft some kind of policy?

Should ChatGPT be banned?

At this point, an all-out ban on ChatGPT isn’t really necessary, says Alida Meghji, associate partner, digital and emerging technologies at EY Canada.

“I think that these technologies are going to become part of our life, if they haven't already,” she says — plus, this is a tool available to anyone on the internet.

“Even if the website is banned on a company tool, employees still have access to it.”

In banning them, employers send the wrong message around what these tools can actually be used for, says Meghji, who refers to ChatGPT as “a C-plus student.”

“It's more about ‘How do we actually harness the good that could come out of these technologies? How do we find the right use cases in organizations… to actually help us be more productive, help our employees be more productive?’ That's where I think we really need to be thinking about, instead of banning the tools outright.”

While a ban could make sense on an exceptions basis, says George Waggott of George Waggott Law in Toronto, it also means “that you lose the benefit of having some good information.”

Previously, Canadian HR Reporter spoke with experts about the pros and cons of a tool like ChatGPT, along with implications for HR.

Why a policy for open-source AI makes sense

Waggott cites the opinions of John Markoff, a New York Times tech reporter, who has said the use of these tools will migrate towards a hybrid solution, where they’re not eliminated but they’re also not relied upon exclusively.

“It becomes more about augmented intelligence,” says Waggott. “You still need that human component because you need to try and understand ‘What are the assumptions? What's the data set we're using? What are the inputs we're using? What's the use we can make from it? Are we actually applying it and using the right information with the right situations?’

“I think that's spot on.”

That’s why a policy makes sense in terms of providing “a guardrail,” says Waggott, “so that people realize this isn't a replacement for human thinking — it's an aid, it's a tool.”

It’s important to understand how powerful these tools are, he says, citing the ability to create “pretty complex content.”

Employees may see this as an opportunity to take away a lot of their grunt work, says Waggott, “but it may also be a recipe for people just switching off in terms of being engaged in what they're producing, and being in any way capable of saying this is authoritative or not.”

“That's a big problem,” he says, because mistakes will come, especially with a tool that is giving predictive information.

“We've got that problem of people using it in a manner which they may think it's authoritative when it may not be.”

Employers need to decide if they are going to allow people to use this tool, and hold out the content as being something other than that generated by the tool, says Waggott.

“If we're talking about HR, legal risks, your organization should be saying, ‘If you are going to use a tool like that, to provide that information, you need to attribute to where you got it from…’ What are the questions? Where were we getting our information from? To my mind that authenticity and accuracy question is something that an organization has to sift through.”

The adoption of artificial intelligence (AI) has more than doubled since 2017, though the proportion of organizations using AI has plateaued between 50 and 60 percent for the past few years.

What kind of policy makes sense for ChatGPT?

The policies will depend on the type of organization and employee involved, says Meghji, “to make sure that they can mitigate some of the risks around privacy, security, quality control, etc.”

Employers should be looking to put in place “guardrails,” she says, along with training employees on how the tool is actually used, and how it generates responses — “really making sure that people understand it and understand the risks of putting in information and asking ChatGPT questions.”

“These tools are really new; there's a lot of buzz around them,” says Meghji. “I definitely think that communications and change management is an integral part of rolling out any of these tools and technologies.”

Most employers aren’t using these tools to cut costs — it’s about being more efficient and more productive, having employees take on more strategic and value-add tasks, she says.

“There’s an opportunity to really train teams, really upskill them for what this technology is going to be used for.”

The policies should also cover when an open-source AI tool is being used to create content and whether content has been fact-checked, says Meghji.

‘Lean’ policy works for now

Now is a good time to create some kind of policy around ChatGPT and similar tools because they are becoming a component of a lot of people’s work, says Waggott, and “because it has the appearance and the presentation and the polish authority — it's not apparent that it's got flaws.”

“The reason to have some form of policy or direction about it is because it is pretty clear that it's prone to be misunderstood or abused… and the mandate of an organization includes being able to provide accurate information.”

But at this point, it can be a “lean” policy that’s relatively simple, says Waggott. That means deciding which company systems and resources are going to be used for these chatbots or generative products tool.

“The second piece is whether you're going to require people to identify its use or attribution. And the third component [concerns]… confidential information and personal information, whether you're going to have either a limitation or a prohibition on that being used when interacting with the tool.”

On top of that, employers should expect to have some level of orientation training and familiarity around the new technology, he says.

Subscribe today for more great content from Canadian HR Reporter

The guidelines could even go into the initial employment contract, clarifying that the employer-employee relationship is based on the employee being paid to do the job, and “not using the chat to do the work for you,” she says.

These kinds of policies are also important when it comes to employees handling private or sensitive information, says Mahil, so they understand the implications and are “really careful about any use of it, especially with respect to your work and your work product.”

There are often intellectual property clauses in employment contracts, saying that anything that's created will be the property of the employer, says Mahil.

“But I think now employers are going to have to put something in that says, ‘When you're asked to create a product, it is your own creation, it is the intellectual property of this company’ — so it's not generated by some random bot that spat out an answer.”