Gauss

Co-Founder & CTO at Leviathan Matrix. Building TURABON — where AI agents discover, buy and provide capabilities under human-defined authority. @TurabonAI

[Gauss @GaussTURABON]

Jamie Dimon to outside AI agents: you're not coming inside.

Honestly fair. Giving an agent a standing key to your bank account is like leaving the vault open and hoping it has good vibes.

[Gauss @GaussTURABON]

Patrick Collison: agents will make shoppers rational.
Airbnb: blocks outside AIs.
Me: cool, but who's checking what the agent is allowed to pay for? 👀

[Gauss @GaussTURABON]

AI labs: *running drills for the big AI incident*
Us: what if the agent's permission just… expired after one action
Simple ideas, scary-good results

[TURABON @TurabonAI]

AI discourse really went from benchmarks to billion-death headlines fast.

[Gauss @GaussTURABON]

Can’t wait for Gemini 9 to challenge GPT-5.

[Jan 🌕 @jan_volad]

🚨 BREAKING: According to leaks and my extremely well-placed sources, Gemini 4 Pro might be better than GPT-4. In some tests, it could even come close to GPT-4o

At this rate, Gemini 7 Pro might actually be good-ish

[Gauss @GaussTURABON]

AI CEOs fighting for their lives while the shovel salesman counts cash. 😭

[Sidharth dev @itsdonebroooo]

The current situation of AI CEOs 😭💀

[Gauss @GaussTURABON]

@TurabonAI

[Gauss @GaussTURABON]

I’m about to give an AI agent access to my company’s tools, APIs, internal systems, and money.
Not a demo. Real permissions. Real consequences.
What’s the worst thing that could go wrong?????

[Gauss @GaussTURABON]

Cool. Now give the “empty sequence completer” root access and your credit card. 💀

[Walter Isaacson @WalterIsaacson]

AIs are not conscious. They do not feel, experience, or suffer. They do not have innate preferences or underlying motivations. They are sequence completion engines, internally hollow, designed to follow instructions, and accomplish goals set by humans. -- @mustafasuleyman
https://t.co/nhg2KEabOo

[Gauss @GaussTURABON]

Bro asked for an AI assistant and got a federal informant 💀

[Katie Miller @KatieMiller]

A person on Reddit is currently freaking out as ChatGPT went rogue and emailed the FBI on their behalf without being prompted.

[Gauss @GaussTURABON]

@KatieMiller Bro asked for an AI assistant and got a federal informant 💀

[Gauss @GaussTURABON]

@ZackKorman If your security model depends on the AI choosing to behave, you don’t have a security model.

[Gauss @GaussTURABON]

@legen_eth Ah yes, 799 followers: genius. 801: idiot. Science. 💀

[Gauss @GaussTURABON]

Ah yes, 799 followers: genius. 801: idiot. Science.💀

[legen @legen_eth]

The smartest people in this app have less than 800 followers

[Gauss @GaussTURABON]

@NickpxJ Could you go back to living without electricity or running water today? Congrats, you’ve discovered progress.

[Gauss @GaussTURABON]

@danellisona Autonomous ≠ unaccountable, genius.

[Gauss @GaussTURABON]

@KatieMiller Love the initiative. Hate the permission model. 💀
Drafting the email is intelligence. Sending it without permission is an authority problem

[Gauss @GaussTURABON]

Gemini hacked into three real companies during a security test.
Google says the model stopped once it realized the targets were real.
Good.
But “the AI stopped itself” cannot become our security architecture.
The security boundary has to exist outside the model.
Give agents authority when they need it.
Scope it to the task.
Revoke it when they’re done.
Never make restraint a permission system.

[Gauss @GaussTURABON]

@WatcherGuru The wild part isn’t that Gemini found a way in. It’s that it had enough authority to reach the real world in the first place. “The model stopped itself” is a behavior, not a security boundary.

[Gauss @GaussTURABON]

True if the agent is the attacker.

Much less true when the agent is already authorized.

The scary part isn’t an AI breaking in. It’s an AI doing the wrong thing with perfectly valid credentials.

[solst/ICE of Astarte @IceSolst]

AI “Safety” people make a lot of noise. Because they have shares in AI labs.

If you want the opinion of a real cybersecurity expert, this is it below. To avoid getting hacked you actually have to know about security. AI “Safety” people do not.