HI, CLAUDE; HELLO, MYTHOS PART 2

YESTERDAY WE BEGAN GLEANING TIDBITS from Professor Paul Taylor’s LRB Diary “Ask Claude.” Today in Part 2, he continues with our introduction to Mythos and Anthropic’s novel idea of A.I. ethics. 

Hello, Mythos. Taylor observes, “The situation was already difficult enough when, on 7 April, Anthropic announced it had developed a new model, Mythos. In many ways Mythos is only an incremental improvement on previous models, but its abilities in coding, reasoning and autonomy are significantly greater.”

He continues, “It is incredibly good at finding weaknesses in existing software systems and alarmingly adept at devising ways to exploit them. The engineers testing it were able to find thousands of security vulnerabilities, with some discovered in almost every operating system and web browser on the planet.”

Lowering the Cybercrime Bar.  Taylor describes, “Anthropic’s announcement of Mythos’s astonishing capacities and their obvious dangers generated a blaze of publicity, as was no doubt intended. (The company is preparing for an IPO later this year.) Thomas Friedman wrote in The New York Times of the fear that malevolent actors could gain access to Mythos: ‘Indeed, this is potentially as fundamental and significant a turning point as was the emergence of mutually assured destruction and the need for nuclear non-proliferation.’ ”

“In truth,” Taylor assures, “Mythos isn’t a superintelligence, and most if not all of the vulnerabilities identified in the report would be familiar to cybersecurity experts…. One danger is that, by automating the generation of the necessary code, it could lower the bar for entry into cybercrime.”

Image from Scientific American, April 17, 2026. 

Enter Project Glasswing. “In their report,” Taylor observes, “the Anthropic team argue that advances in tools tend to benefit defenders rather than attackers. But it obviously helps if the defenders have a lead on the attackers and Anthropic have announced Project Glasswing, a phased release of Mythos, which will first be shared with a consortium of tech companies so they can remedy the weaknesses it finds.”

Anthropic’s White Hat. Taylor describes, “Anthropic’s CEO, Dario Amodei, used to be vice president for research at OpenAI…. By 2021, however, he had decided that OpenAI had abandoned its original mission of delivering AI that ‘benefits all of humanity’ and left to found Anthropic, which would have a singular focus on A.I. safety.”

What a Novel Idea: Principles of Ethics. Taylor observes, “Anthropic’s approach to safety is not to create a list of restrictions and prohibitions that acts as a filter on its models’ outputs, but rather to ensure that the models internalise a set of ethical principles set out in a document the company calls the ‘constitution’. In both November 2024 and July 2025 it felt obliged to augment this strategy by building explicit constraints into the contracts with the DoD requiring that its products not be used for the mass surveillance of US citizens or in fully autonomous weapons.” 

Taylor continues, “At some point between July 2025 and the abduction of Nicolás Maduro in January 2026, the DoD’s thinking changed. The problem seems not to have been the specific restrictions, but the fact that a private contractor wanted to set limits on what the US government could do. It became clear that the two sides couldn’t work together and the contract was ended.”

Follow the Money. “Almost simultaneously,” Taylor observes, “OpenAI—whose president, Greg Brockman, and his wife last year donated $25 million to Trump’s super PAC—stepped into the breach. There was a brief backlash from OpenAI employees and customers worried that the company was becoming too close to an unpopular president. Katy Perry tweeted a screenshot of her Claude subscription. In response, OpenAI’s CEO, Sam Altman, who has an increasing reputation for duplicity, claimed to have secured pretty much the same guarantees that the administration had said were unworkable with Anthropic.”

“Woke” and “Left Wing Nut Jobs.” Taylor describes, “Had the U.S. government simply decided that it couldn’t work with Anthropic, there wouldn’t be much of a story here, but that wasn’t its reaction. Dean Ball, a former adviser to the Trump administration on A.I., used the word ‘revulsion’ to characterise the administration’s response to the moral imperative behind Anthropic’s corporate mission. Trump described the company as ‘woke’ and ‘left wing nut jobs.’ ”

In Stumbles Hegseth. Taylor recounts, “Pete Hegseth, the secretary of defence, branded Anthropic a national security risk and declared that any company working with it would be banned from US defence contracts, a measure which, if enforced, would pose an existential threat to the company, denying it access to such essentials as Nvidia chips and Amazon cloud services. In practice, Hegseth has only been able to impose a more limited but still significant ban, excluding only Anthropic itself from the supply chain for defence contracts.” 

Court Haggling Already. Taylor notes, ‘No one seems to think that this is justified or legally defensible, but it may take years of legal wrangling before it comes to court. Anthropic tried to get an injunction for the ban to be lifted in the interim, but that required winning in both a federal and a DC court.”

It ain’t over yet, even as the I.P.O. progresses…. Dario Modei, Anthropic’s co-founder. Image by Don Feria/AP from “Anthropic Files to Go Public, Setting Stage for Huge I.P.O.,” The New York Times, June 1, 2026. 

Guess What: Taylor observes, “The Californian federal judge ruled in their favour, arguing that punishing the company for expressing views the administration did not like was a breach of their First Amendment rights. The panel of three DC judges, two of whom are Trump appointees, ruled the other way. The threat of a wider ban, excluding Anthropic from all government contracts, remains, but the company and the Department of Defence have met, and government agencies are apparently keen to get their hands on Mythos.” 

What? The Trump administration seeking out a moral imperative?? Quelle absurdité! ds

© Dennis Simanaitis, SimanaitisSays.com, 2026