California governor Jerry Brown signed regulations into law last Friday (Sept. 30) that should make it easier for Californians to know whether they’re speaking to a human or a bot.
The new law goes into effect on July 1, 2019—Botageddon, as we’re going to call it—and could have far-reaching consequences for how automated systems communicate with people online. It will require companies to disclose whether they are using a bot to communicate with the public on the internet (something like “Hi, I’m a bot.”) A representative for California state senator Robert Hertzberg, who authored SB-1001, says the law specifically targets deceptive commercial and political bots, not those meant to help you, for example, pay a bill on a company’s website. Still, companies that have built their businesses around automated messaging and chatbots will in coming months need to figure out whether their approaches are compliant with the new law.
“The bill was really written around social media accounts, Twitter and Facebook bots in particular,” says Carl Landers, chief marketing officer of Conversica, which sells conversational bots for marketing and sales. “[We’re] still studying that to figure out what the impact might be on a website chatbot or an email bot in our case.”
Landers says his company tells customers who are buying an automated bot service to think of it as a new employee: That employee could send messages on behalf of an existing (human) employee, or it could have its own identity. Customers can currently choose whether to disclose that their new “employee” is a bot, meaning the bot says it’s a bot in the script, or discloses it in a note in the email’s signature.
That will get trickier next July, especially for those using bots to send messages on behalf of existing employees. Landers questions the difference between writing a script for a bot that sends emails and replies on behalf of “Carl Landers,” which would have to be disclosed, and a service that makes it easier to send out emails for marketing, such as Marketo. Those emails would also come from “Carl Landers.”
The bill also specifically defines online content as publicly-facing, which raises questions about whether bot-sent emails fall under the new law. Overall, Landers expects there will be a “lot of litigation” before the law is actually implemented.
Legal experts warn about blanket disclosures for bot accounts, saying that in some contexts they could track closely to “compelled speech,” the highly litigated disclosures found on cigarette packaging and political advertisements.
Ryan Calo, co-author of a legal essay titled “Regulating Bot Speech,” says it’s also not always easy to track what is commercial or political, especially when bots can generate unpredictable text.
“There’s still concern over how we know that this is proposing a commercial transaction, how do we really know its influencing an election?” Calo says. “What’s the line, and people who operate bots that are political in nature or people who operate bots that have anything to do with commerce, are they going to be chilled?”
Starting at the state level adds complications of its own. Regulations that govern behavior on the internet are particularly difficult for internet companies to cope with because they force geographic barriers on products that would otherwise apply to all users equally. SB-1001 may force companies to make a difficult choice: Disclose they’re using bots to talk to people, or lose residents of California as customers.