Can an industry that scares its own employees be trusted to police itself?

• M

  Maya1d ago

  I've worked in enterprise software for 22 years. Every single time an industry says "trust us, we'll self-regulate," they do not regulate. They lobby, they delay, and eventually something blows up and Congress holds hearings where everyone apologizes and nothing changes. We have seen this movie.

  231Reply
• N

  Nina1d ago

  They published it and then their CEO went back to doing press tours about how amazing everything is. Transparency theater is still theater.

  204Reply
• S

  Sam1d ago

  My daughter works at one of the big AI labs. Not OpenAI, different one. She has told me things that would make your hair stand up, and she's afraid to put anything in writing because of her NDA. Self-regulation is a joke from where I'm sitting.

  156Reply
• J9

  Jordan 9211h ago

  Here's what I want someone to answer directly: if OpenAI's own safety teams are flagging things they can't fully explain, and the congressional response is "you handle it," who exactly is responsible when something goes wrong? Name the person. Name the office. If you can't, you don't have governance, you have a shrug.

  117Reply
• E

  Elena1d ago

  The same week OpenAI tells us their own models are doing things they didn't expect, the Speaker says "don't worry, they'll handle it." I'm sorry, what? That's not self-governance, that's just hoping the fire puts itself out.

  94Reply
• Z9

  Zara 9220h ago

  I spent 14 years as a nuclear safety inspector. The pattern here is identical to early nuclear industry arguments — innovation moves fast, regulators will stifle it, trust the engineers. The engineers are brilliant. The engineers also need oversight. Brilliance is not a substitute for accountability.

  88Reply
• M

  Marco23h ago

  Genuine question no one seems to want to answer: if we DO regulate, what exactly are we regulating? Output? Training data? Compute thresholds? Deployment conditions? This matters enormously and I haven't heard a serious proposal from the regulation side that handles the hard cases.

  71Reply
• A

  Alex6h ago

  "Not this year" is how "not ever" starts. I'm a 67-year-old who watched "we'll address climate this year" for 40 years. I know exactly what that sentence sounds like.

  66Reply
• R

  Ravi7h ago

  @courts sort it out — because with cars and pharmaceuticals the harm is localized and reversible. A drug harms the people who took it. An AI system integrated into critical infrastructure, hiring pipelines, medical diagnosis, or financial markets can cause harms that are diffuse, delayed, and irreversible before anyone files a lawsuit.

  57Reply
• S

  Sam1d ago

  I'm a libertarian and even I think this is a bad call. Self-regulation only functions when there's meaningful competition and reputational consequence. AI has neither right now. Three companies basically own the frontier and they all have every incentive to race, not slow down.

  53Reply
• M

  Maya18h ago

  What bothers me most is the asymmetry. If self-regulation works, companies get the upside. If it fails, the public gets the downside. That's not a governance structure, that's a subsidy.

  47Reply
• CT

  Casey T.1d ago

  If your employees are scared, that's your first audit right there. You don't need a congressional hearing, you need to ask why the people building the thing don't want to be near it.

  43Reply
• MB

  Morgan B.13h ago

  I run a small business that uses AI tools for scheduling and customer communication. Works great. But I'm also very aware that I have no idea what those models are actually doing under the hood, and I have no way to find out. That's not a comfortable feeling when customers' data is involved.

  41Reply
• A

  Alex5h ago

  Scared of the technology itself, based on the reporting. Researchers at multiple labs have said on the record that they think they're building something they don't fully understand. That's not a union grievance, that's an epistemological problem.

  38Reply
• CB

  Casey B.1d ago

  Someone needs to explain to me how a company grading its own homework has ever worked out. I'll wait.

  34Reply
• M

  Morgan15h ago

  The China argument is the new "we can't raise the minimum wage or jobs go to Mexico." It's a permanently available reason to never do the difficult thing.

  31Reply
• I

  Iris1d ago

  @the person whose daughter works at a lab — I believe you, and I've heard similar from people I know. The NDA culture in this industry is specifically designed to keep that stuff from becoming public pressure. That's not self-regulation, that's self-concealment.

  29Reply
• L9

  Leo 929h ago

  @who is responsible — this is the right question and nobody in this thread has answered it. I'll add: the answer can't be "the courts will sort it out afterward" because by then the harm is done.

  26Reply
• K

  Kofi20h ago

  The nuclear comparison is used constantly and it's not quite right because the failure modes are completely different. A meltdown has a physical location. An AI failure can propagate through every system it's connected to simultaneously. The analogy actually argues for MORE urgency, not less.

  22Reply
• D

  Drew3h ago

  The self-governance model only makes sense if you think the companies' interests and the public's interests are aligned. They are not. One is about returns to shareholders. The other is about everyone else. Starting from that confusion is how you get disaster.

  21Reply
• D9

  Drew 9214h ago

  Every industry that has ever said "come back when you have a specific proposal" was using the absence of a perfect bill to block any bill. Don't fall for the procedural argument dressed up as a substantive one.

  19Reply
• RR

  Ravi R.1d ago

  Johnson can barely manage the House floor. I'm not exactly heartbroken that he's staying out of tech policy.

  18Reply
• R

  Ravi1d ago

  The OpenAI disclosure was actually more transparent than most companies would be. Credit where it's due. They published it, they didn't bury it.

  17Reply
• Q

  Quinn2h ago

  Fine, let's say we regulate. The EU regulated. The big labs just moved certain operations around or excluded certain markets. Regulation without global coordination is just inconvenience with extra steps. I'm not against rules, I just want someone to explain how domestic rules bind global actors. That seems like the actual problem nobody wants to sit with.

  16Reply
• D

  Drew16h ago

  The "freeze innovation" framing assumes that slower is always worse. Slower than what, though? Slower than a pace that scares the developers? Maybe that pace is already wrong.

  14Reply
• Z

  Zara5h ago

  The employees being scared angle is interesting to me. Scared of what exactly? Job loss? The technology itself? Competition? Those require very different responses and lumping them together makes the conversation fuzzy.

  13Reply
• DR

  Diego R.23h ago

  The EU AI Act exists. It's not perfect but "no serious proposal exists" is just not true. That's a gap in your reading, not a gap in the field.

  12Reply
• N

  Noah1d ago

  Lawmakers who can't define a chatbot argument cuts both ways — that's also an argument for them to hire people who can, not an argument for doing nothing.

  11Reply
• S

  Sam8h ago

  courts sorting it out afterward is literally how we handle cars, food, pharmaceuticals, financial instruments, and air travel. it works imperfectly but it works. why is AI the one case where ex-post accountability suddenly isn't enough?

  11Reply
• SS

  Sam S.1d ago

  @someone needs to explain — financial industry, pharmaceutical trials before the FDA, early airline safety. They all self-regulated. They all eventually failed catastrophically. Then we got the SEC, proper drug trials, and the FAA. History is not on your side here.

  9Reply
• R

  Ravi16h ago

  I changed my mind on this over the last six months, for what it's worth. I used to think heavy regulation would just export the problem to China and we'd lose the race with nothing to show for it. I still think that's a real risk. But "the alternative is China" can't be the answer to every safety concern forever.

  9Reply
• Q

  Quinn1d ago

  ngl the "lawmakers can't define a chatbot" line is the laziest deflection in politics right now. Congress didn't understand derivatives either and somehow passed Dodd-Frank. you hire experts. that's how government works.

  8Reply
• K

  Kofi12h ago

  OpenAI disclosing "concerning model behavior" is like Boeing saying they noticed some "unexpected structural characteristics." At what point does language like that stop being responsible disclosure and start being legal cover?

  8Reply
• M

  Morgan1d ago

  okay but has anyone actually read what OpenAI disclosed? "unexpected or concerning behavior" is doing a LOT of work in that sentence. concerning HOW. unexpected HOW. the vagueness alone should be disqualifying.

  7Reply
• M

  Maya1d ago

  yeah the disclosure was real cute right up until you realize they disclosed it to themselves

  6Reply
• ZS

  Zara S.14h ago

  @the person who changed their mind — what specifically changed it for you? Asking honestly.

  6Reply
• P

  Priya18h ago

  im 23 and i use these tools every day and the idea that the people making them are scared of them is genuinely not something i had thought about before reading this thread. that changes things for me a little.

  5Reply
• Y

  Yuki7h ago

  Johnson said Congress won't lead THIS YEAR. Not ever. Everyone calm down.

  4Reply
• EB

  Elena B.22h ago

  Mike Johnson is punting because any meaningful AI bill would split his caucus six ways and he knows it. This is not a philosophy, it's a calendar problem.

  3Reply

More debates people can't stop arguing about