OpenAI Begged for AI Regulation and Then Lobbied to Water It Down

To follow up on last week’s post on the passage of the E.U.’s AI Act, Time magazine has learned that OpenAI lobbied for significant elements of the E.U.’s AI Act to be watered down to reduce the regulatory burden on the company. During the last month, the CEO of OpenAI met with heads of governments throughout the world, speaking about the need for global AI regulation. In several instances, OpenAI proposed amendments that later made it into the final text of the law. In 2022, OpenAI repeatedly argued to European officials that the AI Act should not include its general-purpose AI systems.

Microsoft, which has invested $13 billion in the AI lab, and Google also lobbied E.U. officials, maintaining that the burden for complying with the Act’s most stringent requirements should not be imposed on large companies that build general-purpose AI systems. Daniel Leufer, a senior policy analyst focused on AI at Access Now’s Brussels office, said, “What they’re saying is basically “please regulate us,” and when confronted with regulation as law, they say “no.” It should be noted in passing that OpenAI had specifically claimed that Chat GPT-3 should not be included in the Act’s high-risk designation. All this despite researchers having demonstrated that Chat GPT can be vulnerable to specific prompts and send phishing emails or recipes for dangerous substances.