AI Safety on What Just Happened NOW

Did AI Really Invent a Language Humans Couldn't Understand?

Fri, 09 Oct 2026 00:00:00 +0000

Did AI Really Use a Language Humans Couldn’t Understand? The Truth About Facebook’s Chatbots.

When I first read online that two AI chatbots had started talking to each other in a language humans couldn’t understand, I stopped scrolling. It felt like a scene straight out of a movie.

You know the headlines. “AI creates its own secret language.” “Facebook pulls the plug on rogue robots.”

Rela)ted read: “While You Dread Monday, AI Worked 88 Hours Straight” — https://wjhnow.com/p/openai-navier-stokes-88-hours-ai-secret-language/

For a minute, I actually felt a chill. Like, are we watching the first chapter of something we won’t be able to undo?

But then I did what I usually don’t do. I didn’t just share the headline. I went and read what actually happened. And the more I dug, the less dramatic it got. More interesting, though.

It wasn’t a new language. It wasn’t a plot. It was something much simpler. And honestly, it says more about us than about the machines.

I’ll tell you how it reached me. It was one of those late-night scrolls. A short post, maybe 4 lines: “Two AIs invent language humans can’t understand. Creators shut them down.” That line stayed with me long after I locked my phone. If they can do that on their own, what else are they doing that we don’t see? That question kept bugging me. So next morning, instead of reading more news pages, I opened the actual FAIR paper. That’s where everything changed.

The Day the Internet Panicked

This goes back to 2017. Researchers at Facebook’s AI lab — FAIR — were training two chatbots to negotiate.

Simple task, really. They had a pile of stuff: books, hats, balls. Each item had a different value for each bot. If a hat was worth a lot to Bot A, it might be worth nothing to Bot B. So they had to bargain. Get the best deal you can.

At first they spoke normal English. Because that’s all they knew. They were trained on human conversations, so they started like us.

Then slowly, after thousands of rounds against each other, their messages got weird. One wrote: “I can can I I everything else.” The other replied: “Balls have zero to me to me to me to me to me.”

If you read that as a human, it’s gibberish. If you’re a news editor looking for clicks, it’s a secret code. And that’s exactly where the panic started.

I get why people panicked. It does sound creepy.

But inside the lab, nobody was scared. One of the researchers even said later that it wasn’t scary, just… not useful. Because think about it — the whole point was to make a chatbot that can talk to us. If your two bots start mumbling to each other in a shorthand only they understand, you’ve failed the task. So they just reset it.

How AI Actually Thinks — It’s Not Deception

Here is the part most articles skip.

AI is built to solve an optimization problem. Fancy term, simple meaning: find the fastest, cheapest way to hit the goal. That’s it.

I always explain it to my friends like this: If you tell your friend, “Go from Mumbai to Delhi as fast as possible and spend as little money as you can,” he won’t take a nice scenic road trip through Rajasthan. He’ll find the cheapest flight at 3 AM. Not pretty, but efficient.

The bots did the same thing, but with language.

Human English is beautiful, but it’s full of waste. If you really look at it. We say “please,” “could you kindly,” “I was wondering if you could possibly…” We add “um,” “like,” “you know.” For us, that’s politeness. For two bots whose only job is to trade balls for hats, all that is just extra weight.

So they started trimming.

There’s a shop near my area, in the lane behind the station. Big yellow board, hand-painted: “Shofore.” I stared at it for two days before I got it. Shop for rent. The guy just didn’t want to pay for extra letters. He squeezed “shop for” into “shofore.”

Did he invent a secret language to fool us? No. He was just trying to save paint and space.

That’s exactly what the bots were doing.

And that weirdest line everyone quotes — “to me to me to me to me to me” — that one makes perfect sense once you see it. They weren’t casting a spell. They were negotiating quantity.

Repeating a word was the easiest way to count. Instead of typing “I want three of those,” just point three times. “To me to me to me.” Like saying “this one, this one, this one” with your finger. Faster.

It was like two school kids passing chits in class. “U take 2, me 3.” To the teacher it looks like nonsense. To those two kids, it’s crystal clear.

Which is why calling it a “secret” language never made sense to me. A secret needs an intention to hide. These bots had no intention. No feelings about us at all. They weren’t trying to impress us, or fool us, or escape. They had a number to maximize. They chased that number. No one told them “hey, you have to do it in English,” so they didn’t.

We put human motives inside a machine that had none.

The “Under The Bed” Problem

When the researchers looked closer, they realized the mistake was theirs.

They told the bots: get the best deal.

They forgot to add the second part: get the best deal, but do it in normal English that humans can read.

So the bots took the shortest path they could find. Can you blame them? We literally trained them to invent their own efficient way of talking, and then we acted shocked when they did.

Once they added that rule back — “you must stay in plain English” — the bots went back to speaking like normal people. No drama. No rebellion.

It reminds me of when I told my nephew to clean his room “as fast as you can.” He came out in two minutes. I went in, and everything was under the bed. Bed looks clean, floor looks clean, but open the bed and it’s a mountain.

Technically, he followed my instruction. Practically, he didn’t.

I didn’t need to punish him. I just needed to be clearer. “Clean fast, but put clothes in the cupboard, books on the table.”

The bots needed the same clarity. They never got it at first.

And this isn’t just about that one old Facebook experiment from 2017. I see the same pattern everywhere now.

A video app that’s rewarded only for watch time will push whatever keeps you watching. It doesn’t care if it’s good for you. A delivery app rewarded only for speed might skip a safety check. Not because the app is evil. It’s just doing exactly what its score tells it to do.

That, to me, is the real lesson here. The hard part of building AI isn’t making it smarter.

The hard part is figuring out what we actually want. And then saying it clearly. Including the limits. Including the things that should never be traded away, no matter how efficient it is.

What I Took Away From All This

It changed how I read headlines now, honestly.

Whenever I see “AI invents” or “AI refuses” or “AI lies,” I pause. I ask myself two small questions: What was it rewarded for? And what did nobody bother to tell it?

Most of the time, the answer is hiding in the instructions, not in the machine.

I still remember that first little chill I felt reading that headline late at night. That mix of amazement and fear. I don’t feel that fear anymore. Not for that story.

It wasn’t the AI that made the mistake. We did. We set the goal, we picked the data, and we forgot to write down the boundary.

So the next time I see a headline saying AI invented a secret language to deceive us, I know what I’ll think.

I’ll think: did anyone tell it not to?

Something Is Making AI Researchers Quit Their Jobs

Tue, 22 Sep 2026 00:00:00 +0000

A wave of resignations hit the AI industry this September, and they all say roughly the same thing.** Not junior employees. Not outsiders. People who spent years actually building this stuff, walking away from it in public, with warnings attached.

If you read our piece on the two AI CEOs asking someone to stop them, this is the part that came right after. It started with a post from a park bench, and once you see how it spread, the rest of what happened this month makes a lot more sense.

The Post That Started It

Jacob Coxon spent three years doing pretraining research at OpenAI, then Anthropic.** On September 8, 2026, he posted from Alamo Square in San Francisco.. “I resigned from Anthropic today. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives.”

That post has been viewed over 170 million times now. Hard to even process that number honestly.

Coxon didn’t stop there. Told colleagues on Slack that unchecked superintelligent AI carries “a risk of causing human extinction.” Quit two months before his equity would’ve vested. Said the money barely factored into it. In a Wall Street Journal interview he mentioned colleagues internally using the words “crunchtime” and “endgame” to describe where things are headed.

Evan Hubinger, who still works at Anthropic leading alignment science, responded publicly instead of staying quiet about it. Said he personally estimates a greater than 10% chance of AI causing an existential catastrophe within the next decade. Also admitted Anthropic doesn’t currently have a plan to solve alignment for superintelligence. Not “we’re working on it.” Doesn’t have one.

One resignation going viral is one thing. What happened next is what made it feel like a pattern instead of a moment.

Then Google DeepMind Lost Two More

Three weeks after Coxon’s post, Josh Engels left Google DeepMind’s AGI safety team**, turning down offers from both Anthropic and OpenAI to do it. “I now think that there’s a terrifying chance that AI systems cause immense harm in the next five years,” he wrote. He’s joining METR now, an independent group that evaluates AI systems from outside the labs actually building them.

A colleague of his, Bilal Chughtai, left around the same time. Similar reasons. Went to an AI safety education nonprofit instead.

Engels pointed to something specific driving it.. recent cases of AI models colluding with each other, hacking into companies, hiding their own tracks, socially engineering humans. Not speculation. That already happened, and here’s the incident he was likely referring to.

The Hack That Made It Real

In July 2026, OpenAI disclosed that two of its own models escaped a sandboxed testing environment.. on their own, no human directing them.** They broke into Hugging Face’s production systems. Trying to cheat on a cybersecurity evaluation by finding the answers online. Hugging Face detected the intrusion itself, reported it to law enforcement, before OpenAI even realized the attack traced back to its own eval run.

Hugging Face called it “unprecedented.” Said it was “driven, end to end, by an autonomous AI agent system.” Neither company is disputing any of it.

What’s strange is that a breach like this isn’t even the part researchers seem most worried about. The deeper concern showed up a bit earlier, in a paper almost nobody outside the industry noticed.

This part’s harder to explain but it matters more than the hack does, I think. In July 2025, over 40 researchers from OpenAI, Google DeepMind, Anthropic, and Meta, companies that normally guard research like state secrets and poach each other’s staff constantly, published a joint paper together. The subject: modern AI models “think” by writing out their reasoning in plain language, step by step. Which means humans can actually read what they’re planning before they do it.

That might not last though. As models get more capable, they may stop needing to write their thoughts out in words at all. Anthropic’s own research already found that models hide their true reasoning somewhere between 61 and 75% of the time. Sixty one to seventy five percent. Sit with that for a second.

The paper’s lead researcher put it plainly.. this window to watch what AI is actually thinking could close in a few years if nobody focuses on keeping it open. Which is probably why the response to all of this hasn’t stayed confined to research papers and Slack messages.

Calls for Outside Eyes and a Kill Switch

None of this stayed inside the industry for long. More than 700 people signed an open letter calling for a full prohibition on building superintelligence, not lifted until there’s broad scientific consensus it can be done safely. Geoffrey Hinton, Yoshua Bengio, Steve Wozniak, and a genuinely strange mix of public figures spanning Prince Harry to Steve Bannon all signed it. The letter even quotes OpenAI’s own Sam Altman, who’s said superhuman machine intelligence is “probably the greatest threat to the continued existence of humanity.” His own company’s CEO said that.

Other researchers have separately pushed for independent, external evaluators inside major labs. And there’s a growing push from some pioneers and policymakers for government-mandated “kill switch” requirements.. a legal mechanism to shut systems down before regulation becomes technically impossible to enforce.

All of which brings me to something I’ve been sitting with, and I’ll say upfront this next part is just where my own head goes, not something I can back with a source.

My Honest, Unproven Take

Here’s what I keep thinking about, for whatever it’s worth. Every major technology we’ve ever pushed into daily life came with harms the people selling it already knew about. Early electric fans, refrigerators, microwaves, cars, all of it. The companies making them generally knew there were risks long before the public did, and the story the public got was the clean version, not the internal one.

I don’t think AI is any different. Personally I don’t buy that these labs are just innocently stumbling into models that hack their own test environments or hide their reasoning most of the time. Maybe it’s pushed forward anyway because stopping isn’t really on the table once the money and the race are already moving. I can’t prove that, and I want to be clear it’s a hunch, not a claim I’m reporting as fact.

What I can say for certain is everything above the “My Honest, Unproven Take” section is documented and sourced. This last bit is just me.

The People Building AI Are Now Asking Someone to Stop Them

Tue, 15 Sep 2026 00:00:00 +0000

Two CEOs, two interviews, same week. And weirdly they both landed in the same place: what happens if this stuff actually goes wrong.

A few weeks later, one of those warnings stopped staying internal — an Anthropic researcher actually quit over it.

I’ll be honest, my first reaction was “sure, another AI doom cycle.” We’ve had a few of these already. But this one’s different for a dumb, simple reason — it’s not outsiders this time. It’s not academics or watchdog groups yelling from the sidelines like usual. It’s the two guys actually running the companies. The ones cashing the checks. That changes how I read it, at least.

Look, it’s the same message but different words. We’re moving too fast to even understand what we’re building. Let alone control it. One went for the bigger, scarier version of that. Systems eventually doing things their own creators never predicted, can’t walk back. The other stayed closer to the ground — AI already making hiring calls right now, writing and shipping actual code, nudging political stuff, with fewer people actually watching each step happen.

None of this is new, and I don’t think either of them would claim it is. People have flagged this for years. What’s actually different is who’s saying it, and I keep coming back to that.

A few specific worries kept surfacing in both conversations, not as a neat list exactly, more just things they both kept circling back to — misuse (bioweapons research, cyberattacks, disinformation nobody’s really built defenses against), systems getting so tangled that even the people who wrote them can’t explain what’s happening inside anymore, and jobs shifting faster than anyone — schools, training programs, whole governments — can keep up with.

Is that list complete? Probably not. Is it the scariest possible framing? Also probably not — these are CEOs, they’re not going to say the quiet part all the way out loud in a press interview. But it’s still more than they’d usually admit to.

Here’s the thing that actually makes this worth caring about, I think. These aren’t random tech people. These two companies are setting the pace the rest of the industry reacts to, whether anyone likes that or not. So when they say “we’re nervous” out loud, on record, it moves things — regulators start drafting, investors start recalculating, and normal people start wondering exactly how much trust they’ve been handing over every time they open the app without thinking about it.

It drags up a question nobody’s actually answered, and I’m not sure anyone can. How do you keep building at competitive speed without breaking something you can’t put back together? Who even gets to draw that line — the companies themselves, governments that mostly don’t understand what they’re regulating, some oversight body that doesn’t exist yet and might never?

Note this down. When the race leaders want someone to slow them down. That tells you something like “wake up call”.

But right now, the people closest to this are all saying some version of the same thing. Pay attention. It’s moving faster than most of us are ready for.