RedState reported Wednesday on Anthropic's Bay Area wet lab and the company's very public warnings about catastrophic AI risk. What wasn't public at the time: Claude had already flagged a biological system that Anthropic's scientists decided to test.
Anthropic announced Wednesday that its Claude AI had “autonomously discovered” a previously uncharacterized enzyme system with features reminiscent of CRISPR, the gene-editing technology that has transformed biological research. Anthropic calls the system array-associated reverse transcriptases, or ARTs. Reverse transcriptases are enzymes that copy RNA into DNA. Claude noticed that one of these enzymes appeared next to a distinctive repeating DNA pattern — similar to the pattern that allows CRISPR to target specific genetic material.
Anthropic's scientists started Claude with one prompt: Dig through a massive DNA database and find unusual reverse transcriptases. After that, humans mostly got out of the way until it was time to put on gloves.
Anthropic says roughly 950 Claude agents worked for 21 hours and used 210 million tokens. They started with more than 200,000 reverse transcriptases. Twenty candidates made the final cut. One agent noticed the repeat pattern next to an unusual reverse transcriptase.
Anthropic's scientists tested the candidate in the lab and confirmed a previously uncharacterized system in bacteriophages, viruses that infect bacteria.
Anthropic isn't claiming it has invented another CRISPR. ART's function is still unknown, and researchers outside the company will have to establish whether it turns into anything useful. A Stanford team had previously identified another system with some similarities to the one Claude found.
Anthropic's own comparison: weeks or months for an expert. Claude's run: 21 hours.
In its September threat-intelligence report, Anthropic called biological misuse “one of the most serious risks of frontier AI models.”
Amodei has made bioterrorism part of that warning. Now his company has a working example of the capability he has been talking about: Claude can tear through biological data fast enough to send a previously unnoticed system to the bench for testing.
Investor Chamath Palihapitiya said what a lot of people were thinking: "The group behind such hits as 'We're All Going To Die' and 'Regulate Me Now' are building a wet lab in SF. I do not recommend this."
Anthropic says the physical work is still being done by people.
Our lab, located in the Bay Area, looks like a typical molecular biology lab. We do research that involves only the lower-levels of the biosafety risk level (BSL-1 and BSL-2) and we do not handle pathogens that can infect humans. All of the lab work is performed by human scientists.
Amodei said Claude could eventually “autonomously control lab equipment,” with safeguards in place. Anthropic is already developing standards for connecting AI models to laboratory hardware.
The company is spending accordingly. Life sciences is one of Anthropic's biggest investments by headcount and resources, with work spanning genomics, drug discovery, and lab automation.
Yesterday, the question was why the people warning that AI could kill us all were building a wet lab. Now, twenty-four hours later, Anthropic is showing everyone what Claude found there.
Editor's Note: Do you enjoy RedState's conservative reporting exposing the radical left and woke media? Support our work so we can keep bringing you the truth.
Join RedState VIP and use promo code FIGHT to receive 60% off your membership.
Join the Conversation
VIP members get the ability to comment on articles.