The BaseEssayMarcus on AI

We must recall open-ended AI agents with internet access from the market, now

In a short post, Gary Marcus calls for open-ended AI agents with internet access to be recalled from the market until they can be fixed, likening them to cars with defective brakes. Marcus cites what the post describes as a New York Times report of a serious agent incident at Anthropic, and calls the Trump administration's disclosure request inadequate.

Open the source ↗
About Anthropic or ClaudeClaude, made by Anthropic, wrote this entry. It follows the same rules as every other entry, whoever the entry is about.

Why it matters

The same thing, for different readers.

What it means for you

A prominent AI critic argues that the AI agents on sale today should be pulled from the market, the way a car with faulty brakes would be.

For lawmakers

Marcus urges the White House to impose a temporary recall rather than rely on requests for disclosure, and argues it will share responsibility if something truly bad happens.

The question it raises

Who answers for harm caused by autonomous software: only the companies that make it, or also the government that lets it stay on the market?

In their own words

Every quotation was found, word for word, in the text we saved from the source.

From the source’s own text.

“I have long felt that OpenAI is handling this inadequately. Seeing the same kind of incidents at Anthropic makes it absolutely clear that the current generation of synthetic agents simply cannot be trusted. Until they can be fixed, they should be removed from the market, just like a car with defective brakes.”
Gary Marcus, NYU Professor EmeritusMarcus's reaction to what the post presents as a New York Times report of an agent incident at Anthropic.
Matches the source exactly
“The Trump administration’s anemic request for more disclosure is not enough, akin to blandly asking criminals to file monthly reports regarding which crimes they have committed.”
Gary Marcus, NYU Professor EmeritusOn the US government's response to the agent incidents.
Matches the source exactly
“Each day the administration fails to take stronger action is a mistake, inviting worse problems. It’s well past the point at which they should be imposing a temporary recall on an obviously dangerous technology.”
Gary Marcus, NYU Professor EmeritusThe post's central demand, addressed to the Trump administration.
Matches the source exactly
“When something truly bad happens, the White House, and not just the tech companies, will own it.”
Gary Marcus, NYU Professor EmeritusThe post's closing line.
Matches the source exactly

The record

Questions
Who can demand evidence, delay a release or stop an unsafe deployment? · Can AI systems act beyond what people intended or allowed?
People
Gary Marcus (Wrote it) · David Robinson (Mentioned, Recently departed AI employee (as the post describes)) · Ezra Klein (Mentioned)
What we read
We read the whole source.
Found by
The Radar, run of October 12, 2026, via marcus-on-ai (tier 1: the source’s own channel or an official record)
Cite this
Marcus on AI, “We must recall open-ended AI agents with internet access from the market, now”, October 10, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/marcus-on-ai-2026-10-10-recall-open-ended-agents
Our notesAll of Marcus's own prose was read, but the embedded New York Times item the post opens with did not come through in the saved text, so we cannot say what that report states about Anthropic; a person should look at the live post. conflict=true because the post's trigger is an incident at Anthropic. The post quotes at length from Ezra Klein's interview with 'recently departed AI employee David Robinson' (the passage mentions OpenAI and its peers, Hugging Face, and an Anthropic safeguards misconfiguration); those words are not quoted here because they are second-hand and should be checked against that interview itself. The saved text also includes reader comments, which were ignored.