Well, as you and I thought, the digital communism shop is already closing, the corporate watchdogs have woken up, but there is still hope for a revolution.
404 Media published a great analysis of how OpenAI and Anthropic started massively banning and throttling people for reverse engineering. The last two weeks have been the best use of artificial intelligence: guys pitted Claude and Codex against their devices' firmware, cut tons of advertising junk from Xiaomi and LG TVs, removed banners from video babysitters, wrote a custom operating system for old digital cameras, and revived smart mirrors that manufacturers turned into server pumpkins after server shutdowns.
But the celebration of life did not last long.
The most revealing case happened to a security guy under the alias log4j. Someone bought a new Samsung TV, was amazed by the overly accurate advertising and decided to tinker with the firmware using AI.
Inside was a camp of marketing spies - the TV not only captures the screen, but also activates a local neuron that listens to the room for crying children or a barking dog, and even tries to use external sound bars as microphones.
A few days after the analysis was published, OpenAI quietly destroyed the researcher's account and froze $500 in the account for "cyber abuse."
When there was an outcry on Twitter, OpenAI started making excuses: their employee said they apparently didn't ban reverse engineering, just "complex security algorithms accidentally worked", and after 11 days the hacker himself returned his account with a squeak.
But Anthropic followed the principle - their Claude now firmly cuts off any attempts to crack the code and refuses to analyze even local Windows blue screen plots, sending ordinary users to get special permission as for enterprise test.
The irony here, of course, is cosmic. Corporations can feed their models all over the internet and other people's code without asking, but when you try to use the same tool to remove surveillance from a TV you bought for your apartment with your own money, you're instantly a dangerous cybercriminal.