Hello again, and welcome back to Fast Company’s Plugged In.
For years, progress in AI has been motivated by an industry-wide yen to create software that’s at least as capable as humans—not at some tasks, but all of them. The precise definition of the goal varies, and two maddeningly overlapping terms, artificial general intelligence (AGI) and superintelligence, both get bandied around. But no matter how you look at the aspiration (or how long you think it will take to achieve), it’s about the ways the world will change when software can do everything extraordinarily well.
I’ve written—here and here—about why I believe fixating on that eventuality isn’t the best way to think about AI and its impact. It might turn out that AI trounces humanity at some jobs and never rivals it at others. That would not be reason to take it any less seriously. This week brought some of the clearest evidence of that point so far.
On April 7, Anthropic announced a new version of its Claude model called Claude Mythos Preview. Like existing Claude versions such as Sonnet and Opus, it was trained for general competency, not to be a specialist at anything in particular. But Anthropic says that when it tested Mythos, it discovered it had made dramatic strides in coding ability. It was particularly good at finding and exploiting vulnerabilities in existing software, surpassing “all but the most skilled humans.”
According to Anthropic, Mythos detected security flaws in every major operating system and web browser. It spotted a 28-year-old hole in OpenBSD, an operating system designed, above all, to be secure. It also found a 16-year-old one in a widely used piece of video software called FFMPEG that had gone unnoticed even after 5 million rounds of automated testing.
As impressive as that sounds from a technical standpoint, it’s also deeply unsettling. Rogue nation states, low-rent scammers, and other bad guys have long exploited bugs to carry out attacks. Until now, the supply of such flaws has been limited by human ability to uncover them. If AI can perform that work with unprecedented aptitude, anything that runs on software would be radically more prone to attack, from your smartphone to the country’s electrical grid.
Just to make matters more unnerving, Anthropic says early versions of Mythos behaved in various “reckless” ways, sometimes when prodded and sometimes on their own initiative. When the model was isolated in a sandbox that theoretically denied it internet access, it figured out how to break free and send one of its researchers an email. It also made changes to code and then covered its tracks, as if it was hiding something.
Overall, according to Anthropic, Mythos behaved more responsibly than the current Opus 4.6 model. Still, its deep understanding of software vulnerabilities, determination to achieve its goals, and apparent willingness to be sneaky do not sound like a great combination.
Anthropic, which wears its dedication to AI safety like a badge of honor, is moving gingerly. Instead of making Mythos publicly available in its current form, the company has launched an initiative called Project Glasswing to carefully share it on a need-to-know basis. Forty technology companies will get access to the model and a total of $100 million in usage credits, including big names such as Amazon, Apple, Cisco, Google, Microsoft, and Nvidia. That will give them a potentially transformative new tool for identifying and patching holes in their own products. Allowing them to poke at Mythos should also clarify whether Anthropic’s own lofty assessment of its model’s cyberhacking prowess is at all hyperbolic.
Anthropic’s archrival, OpenAI, isn’t part of Project Glasswing. However, Axios’s Sam Sabin reported on April 9 that OpenAI plans to take a similarly cautious approach with an upcoming cybersecurity-savvy version of its own GPT model. Other AI overlords, such as Amazon, Google, and Microsoft, may follow suit.
But no matter how responsibly Big Tech behaves, Mythos surely foreshadows less carefully guarded AI acquiring similar skills, possibly within months. Open-source models could give bad actors uncontained hacking superpowers. Governments would have every incentive to invest heavily in the technology and put it to shadowy use. Cyberterrorism might evolve from looming threat to terrifying everyday reality. Even well-meaning uses of the technology could go awry if a model misbehaves, either by fluke or intent.
The bottom line: Project Glasswing isn’t just about preparing the world for Mythos. It’s also a first pass at readying the tech industry for Mythos-like models that have few if any safeguards, or that are explicitly designed to wreak havoc.
Unsettling though all this is, I find strange comfort in the fact that the tech industry is being forced to face its implications right this very minute. Predictions of when AGI might arrive vary wildly, even among people whose expertise is unimpeachable: “Before the Trump administration ends” and “sometime in the 2060s” are both defensible answers. With no consensus on how much time we have to gird ourselves, it’s tough to make a plan. True AGI would also present us with so many new challenges on so many fronts that confronting them all in parallel would be overwhelming in itself.
Tackling the societal catastrophes that AI may unleash one problem at a time sounds far less intimidating than playing Whac-a-Mole with all of them at once. Now is the best time to start. And if this new age of cyber insecurity turns out to be among the most daunting of the lot, we should consider ourselves lucky.
You’ve been reading Plugged In, Fast Company’s weekly tech newsletter from me, global technology editor Harry McCracken. If a friend or colleague forwarded this edition to you—or if you’re reading it on fastcompany.com—you can check out previous issues and sign up to get it yourself every Friday morning. I love hearing from you: Ping me at hmccracken@fastcompany.com with your feedback and ideas for future newsletters. I’m also on Bluesky, Mastodon, and Threads, and you can follow Plugged In on Flipboard.
More top tech stories from Fast Company
The U.S. and Silicon Valley may be running out of time to deal with Taiwan
Most advanced chips, including those for AI, are produced on the island, which China has long believed to be its own. Read More →
How GoPro lost its way
After years of missed bets and shrinking relevance, the once high-flying camera company is cutting staff, losing money, and nearing a potential delisting. Read More →
Inside Niantic Spatial’s audacious plan to scan the world
The company behind ‘Pokémon Go’ is now creating an AI- and robot-ready 3D model of everything around us. Read More →
Satirizing Silicon Valley is pointless in 2026. This show proves it
AMC’s ‘The Audacity’ is a sharply written, well-acted new series about the foibles of Silicon Valley’s movers and shakers. It’s also utterly inessential in our current era. Read More →
Lawmakers want to restrict 3D printing to stop ghost guns. Critics say it won’t work
New ‘print blocker’ proposals would force printers to scan and reject files, a move opponents say risks privacy and targets the wrong problem. Read More →
These heat pumps can be installed in an hour—and will cut costs in half
Merino Energy, a new startup led by a former Apple engineer, created a lower-cost heat pump that launched today. Read More →
Source link
