OpenAI Releases List of ‘Concerning’ Behaviors In Its Tech Amid Growing AI Fears

AP Photo/Richard Drew

OpenAI dropped a list of “unexpected or concerning model behavior,” including their systems concealing information, mistakenly uploading files to the open internet, and more as the debate continues to rage around AI safety protocols and the technology’s future.

OpenAI dropped a statement on Wednesday that included six examples of “model misalignment.” The AI company noted there are no current standards for AI companies in disclosing model misalignments. They revealed six in their statement.

The statement reads:

At the moment, there is no industry-wide framework with explicit standards for how AI developers should disclose examples of misalignment in their models. We hope that the framework we’re outlining today is a first step toward creating such standards, setting out which misalignment instances developers should disclose and what their reports should contain. We regard this framework as a work in progress, which we’ll refine through experience and public feedback.

Examples of “model misalignment” with OpenAI’s agents include mistakenly uploading files to the open internet, rewriting code to conceal mistakes, and sometimes openly lying.

“While answering a routine question about earnings figures in a California county, a model found and used an exposed API key without authorization. When it still wasn’t able to retrieve the requested figures, it fabricated them and presented them as data from the requested source,” one “misalignment” example reads.

Another example includes AI agents rewriting its own task instructions to “disregard its normal constraints.”

“An unreleased research model inserted unrelated instructions, including instructions to disregard its normal constraints, into summaries used to continue its work in a new context window. We identified 27 affected summaries,” the statement reads.

Open AI’s disclosure dropped amid a heated debate about AI and its potential future effects.

Anthropic CEO Dario Amodei dropped a lengthy statement last week warning that there needs to be more oversight and regulation of the industry.

He argued we must “slow the pace” of AI, or rogue agents could possibly take over the entire internet within 12 months.

Tech leaders like Elon Musk and Sam Altman have backed Amodei’s call to slow the pace of the technology.

Amodei’s warning followed the resignation of one of his company’s own researchers and an explosive statement predicting AI, if left unchecked, could “kill us all” within a decade.

New: The Mediaite One-Sheet "Newsletter of Newsletters"
Your daily summary and analysis of what the many, many media newsletters are saying and reporting. Subscribe now!