As more enterprises seek assurance that the AI tools they use won’t fail or produce undesired effects, AI voice vendor ElevenLabs rolled out an insurance policy for AI agents.
The voice AI lab, a 2022 startup, on Feb. 11 said it secured the Artificial Intelligence Usage Certification-Level 1 (AIUC-1) certification, a process that audits an AI system’s ability to address risks such as data and privacy, safety, security, reliability, accountability and societal impacts. The vendor said that, based on the test results, insurers can now offer an AI-specific insurance policy that underwrites the actions of AI agents deployed by ElevenLabs’ customers. However, it is unclear which companies are willing to underwrite the policy.
ElevenLabs’ move to insure its agents is a response to enterprises’ concern about risks posed by AI technology and their desire to take more measured risks. Despite the fact that many AI vendors, including Adobe, Google, IBM and AWS, offer indemnification, a legal concept under which one party agrees to pay another for losses, damages, or liabilities, insurance for AI agents might help take the protection one step further because it promises a safeguard against what the AI agent itself does wrong.
“This is a reflection of where we are in the enterprise today,” said David Nicholson, an analyst at Futurum Group. “We moved from the fear of missing out to the fear of messing up.”
What Insurance Means
He added that enterprises now routinely weigh whether implementing a given tool will cause damage, unlike a few years ago, when there was a push to implement AI technology regardless of potential for harm.
For, ElevenLabs, which has seen its technology used maliciously in the past (for instance, the vendor’s technology was used in a robocall incident in 2024, in which the voice of President Joe Biden told New Hampshire residents not to vote in the Democratic primary), insuring its agents means it has taken steps to ensure its technology is safe to a certain extent, Nicholson said.
“They’re confident that if it is used correctly, according to all of the warnings on the label, then it’s not going to do bad things,” he said. “That is not trivial.”
However, insurance for AI agents can also foster a false sense of confidence, as even if AI companies offer insurance, it is hard to prove that the damage was not due to user error.
A Better Way
Nicholson noted that indemnification guarantees that even if a bad actor uses the AI tool in a way it is not meant to be used, enterprises are protected. However, it’s not clear how effective indemnification policies are.
Moreover, an insured AI agent is not high on enterprises’ list of concerns, said Lian Jye Su, an analyst at Omdia, a division of Informa TechTarget. He said enterprises are more concerned with the performance and accuracy of the AI agent.
“It’s really about the enterprise’s ability to deploy the agent with the right AI models and with the right cloud infrastructure and then being able to feed the AI with the right contacts using the internal data,” Su said. “You probably want to address all these foundational challenges before you jump into getting an AI agent that has an insurance policy.”
However, an insured agent might appeal to enterprises in highly regulated industries, such as the financial sector. It is also a way for the text-to-voice vendor to differentiate itself in a market where it competes with many vendors, including OpenAI.
