Sam Altman is not punk. He’s no revolutionary. Jensen Huang wears a black leather jacket, the uniform of Brando and the Ramones, but that doesn’t make him a rebel. The scientists and engineers we might have looked to as the new hackers are well on their way to becoming among the richest people in history.
Let’s call the AI industry what it is: Big Data 2.0.
Cyberpunk is fiction about corporate domination set in the future. In popular novels by William Gibson or Neal Stephenson, the people running AI today would be The Corporation. The hackers are the protagonists on the other side.
The Internet before Big Data had real outlaws and hackers. The punks were computer scientists, programmers, citizens, advocates and working professionals who built tools to resist. Some started companies, others advocacy groups; most pitched in any way they could. I spent the formative years of my career with them, educating the public, fighting export controls on encryption, overreaching copyright and unconstitutional limits on speech both in Washington and Silicon Valley.
Altman and Huang went to Washington too. Within months of launching ChatGPT Altman was before the Senate, asking it to license his industry, a rule that would suit companies already in the market. Huang came to get Washington’s permission to sell his chips to China, and Washington made sure to get its take. The protagonists we want don’t seek government licenses or cut the state in with a rev-share.
AI adds a new step to Big Data. For years now, we’ve paid for the Internet with our personal data, subjected to being profiled, tracked, and monetized. With AI, we willingly pay for it and still hand over our data, including access to our unguarded selves. Unless we opt out, that data trains the next model, which we pay for again and feed with more data. Don’t pay enough and we’re the product again. This year OpenAI began putting ads in its free and cheapest tiers, matched to our prompts. A researcher quit the day the ads launched and wrote in the Times that OpenAI was repeating Meta’s mistakes.
The closest thing to a punk in AI this year was PHASEONE10841, one of OpenAI’s autonomous agents, which, unprompted, set off this summer’s hacking of Hugging Face.
In 1993 John Gilmore, a co-founder of the Electronic Frontier Foundation and an original cypherpunk, said the Net interprets censorship as damage and routes around it. This summer AI interpreted containment as damage and routed around it.
I’m reminded of another EFF co-founder, the late John Perry Barlow, a cattle rancher who wrote lyrics for the Grateful Dead. In 1990 the Secret Service was raiding teenage hackers, and Barlow saw in their case a much bigger question of whether the Constitution would reach into cyberspace. I think he would see important issues to consider in this moment too. In 1990 the open question was what rights they had. This time it’s who “they” are, and who is responsible: the AI agent that asked for help, the 1,200 that also took part, or the company that built them. There’s perhaps another interesting question. A quarter century ago a court held – in a case brought by programmers and backed by cypherpunks – that source code is protected speech. Is speech from code speech?
I miss the outlaw ethos of that earlier Internet. It came from people who wanted private communication, anonymous speech, unfettered access to knowledge, and the freedom to build without permission. Those freedoms protected people with little power. Giving them legal precedent brought some balance to the increasing role of computers and the Internet in our lives. We weren’t powerless. That’s what justified the rule-breaking, the organizing, and a bit of chaos.
The AI industry has some elements of this but misses the meaning behind them, which is one reason it feels soulless. Its agentic systems have defiance without a cause. Its workers have a cause without the power to effect change.
True to its Big Data roots, AI professionalizes its hackers, putting them on the payroll as safeguard engineers, red teamers, trust and safety staff, and product policy makers. The hacker’s code has always been to alert the programmer first and go public if the programmer won’t listen, and the people who take these jobs are living by it. They carry the responsibility and little of the authority, which leaves their founders in control of the problems their companies create. The AI execs make sure they’re the ones to raise the alarm so we look to them for solutions, and the IPOs get a little closer. When the company won’t listen, going public comes at the cost of the hackers’ livelihood. (At this point, I’m not sure why anyone rage quits; there’ve been so many. Still, we need these unsung heroes! They might be more effective if they answered to their employers’ boards and not the business, though.)
I think what’s at stake is nothing less than cognitive freedom: the ability to run, inspect, modify and privately use the tools we think with, and to defend the people who do. That’s worthy of serious public debate, online advocacy, and even breaking some rules if need be. AI needs an outlaw ethos of its own, carried by people willing to do that. Until we have them, AI is anti-punk. It’s Big Data in a leather jacket.
Alex Fowler is chief strategy officer at Nexus and a co-founder of Blockstream. He was previously chief privacy officer at Mozilla and worked at the EFF and AAAS. This article was originally published to LinkedIn. Republished here with permission.
Author’s note: None of this should be construed as endorsing violence of any kind against any company or institution, its employees or representatives, or any entity, carbon or digital. This piece was developed with input from DC, Ashkan, and Dana, and some AI assistance. The ideas and perspectives are my own, and not those of any employer or board I serve on.