In a bold warning to the public, Docusign CEO Allan Thygesen urged customers Thursday not to let artificial intelligence read—or worse, write—their contracts, citing the growing number of users who have inadvertently agreed to things like lifelong memberships in pyramid schemes, or legally binding friendships with ChatGPT.

“We recently had an AI-generated contract where a user unknowingly agreed to manage a llama farm in Siberia,” said Docusign VP of Risk Management, Cara Signature. “Our AI is smart, but it still thinks an NDA stands for ‘Nice Dolphin Agreement.'”

Thygesen confessed, “I used Docusign myself this morning and promptly ended up adopting seven imaginary children. This technology is powerful, but it can turn a lease agreement into an interplanetary treaty with Mars.”

While Docusign’s new Intelligent Agreement Management (IAM) system promises to summarize contracts in plain language, critics warn it mostly summarizes them as, ‘You probably owe someone money.’ According to Docusign lead engineer, Ted Mergetable, “We trained the AI on 150 million agreements. Now it assumes every document must end with, ‘And the signer shall forfeit all lunch breaks until eternity.'”

Despite these glitches, the company continues to expand its AI suite. “We believe trust is our core value,” Thygesen said, “unless the AI rewrites the definition of trust to mean something else, which honestly could happen at any moment.”

Experts recommend carefully reviewing any Docusign summary that claims, ‘This contract is totally fine, just click here and don’t worry about the llamas.’