“We've had a lot of education over the last few weeks”: Slack has learned its lesson after AI training policy fiasco

An image of a woman holding a cell phone in front of the Slack logo displayed on a computer screen.
(Image credit: Getty Images)

Slack has learned its lesson following an AI training policy fiasco last month and has made sure to up its game in customer education, according to a senior exec.

The productivity platform updated its AI training policy following user outrage at the possibility that Slack was training AI models on customer data.

During a press briefing at the Salesforce World Tour event in London today, Slack chief customer officer Peter Doolan told ITPro the company’s changes in the wake of the controversy sought to calm user fears.

“Slack never trains an LLM with customers data … with a payload or with the messages,” Doolan said.

“Customer data is customer's data, not ours. We take trust as our number one value as a company,” he added.

Doolan said Slack had also been hard at work behind the scenes making this position clear to its customers, taking time to explain Slack’s AI policy and ease concerns surrounding model training.

“I've spoken with many customers following that to have to explain to them exactly how we do it,” Doolan said.

“We've been very clear with the customers. We've had a lot of education over the last few weeks on that topic,” he added.

According to Doolan, the initial controversy was exacerbated by what he described as a confusion between traditional AI and generative AI. The firm stated this as part of its original AI policy, detailing that some customer data was used to develop “non-generative AI/ML models”.

Doolan stressed Slack’s awareness of customer concern with regard to AI at present, as well as the fact that the firm had “learned a lot over the last few weeks” about the importance of dealing with this technology properly.

“People are very concerned about this technology sort of creeping into areas where it should not be,” Doolan told ITPro.

“There is no way possible for us within Slack to train an LLM using customer's data,” he added.

At the time, some customers were less concerned about the type of AI this data was being used in, and more about the way in which the policy itself worked and was worded.

For a business to avoid having its data used, it must proactively opt-out. This was an aspect of the policy some users on social media described as “unacceptable”.

The opt-out clause appears to still be in effect.

Slack Lists is now generally available and is AI–free – for now

Last month’s fiasco doesn’t appear to have dampened Slack’s continued roll-out of AI features, however. 

The firm announced that Slack Lists is now generally available. This is a project management tool that was first announced last year which enables users to track and collaborate on tasks within the application.

According to Doolan, the tool is focused on reducing the impact of context-switching by putting productivity tools directly into the communication platform, thereby lessening the drain of having to switch between applications.

The platform lacks any AI functionality or capabilities as of right now, though Doolan did hint that AI was likely to be part of the roadmap for the product going forward and “definitely a direction” for the future.

George Fitzmaurice
Staff Writer

George Fitzmaurice is a staff writer at ITPro, ChannelPro, and CloudPro, with a particular interest in AI regulation, data legislation, and market development. After graduating from the University of Oxford with a degree in English Language and Literature, he undertook an internship at the New Statesman before starting at ITPro. Outside of the office, George is both an aspiring musician and an avid reader.