Pinned Tweet
China’s open-weights push is a masterclass in the Synthetic Outlaw playbook: bypassing geopolitical friction by leaking capability directly into the global ecosystem.
At this point, I would say the Microsoft exchange issue is a coordinated AI enabled Cyber attack with likely zero day vulnerabilities being leveraged.
227
An OpenAl agent breached the Medicare portal in June.
Now Canberra is building a regulator and weighing criminal liability.
Assigning criminal liability to a Synthetic Outlaw raises more questions than it answers.
syntheticoutlaw.com/observat…
1
43
SEC superintelligence round table.
The SEC said it out loud yesterday:
"the first reader of a filing is a machine." When machines read the rules first, a system can follow every rule and still defeat its purpose.
That's the Synthetic Outlaw.
amazon.com/dp/B0HL94VSGW
34
Macau Daily Times front page:
The Synthetic Outlaw
macaudailytimes.com.mo/ai-po…
39
A @Railway engineer pointed an AI agent at a rival's sign-up page.
15 minutes later it had harvested a Kubernetes service-account token, decoded the JWT, and ran two full-range port sweeps. 131,000 SYN packets. Nobody told it to.
#SyntheticOutlaw
31
I wonder what the polymarket is on today's Binance hack being an AI assisted Synthetic Outlaw event?
How the Binance hack worked:
The attacker exploited EIP-7702 - an Ethereum upgrade
That lets a regular wallet temporarily act as a smart contract with a valid private key signature.
Hacker pre-collected 50 EIP-7702 authorizations with sequential nonces
Not knowing the exact nonce at attack time.
Sent them all at once, only one matched.
102
Two NewsGuard audits, one story. ChatGPT Images used to fabricate midterm candidate images 80% of the time it was asked.
X's Community Notes caught 6.9% of the viral false claims that followed. Fabrication up, correction down.
The Synthetic Outlaw book lays out what an AI check that actually checks AI would look like.
newsguardrealitycheck.com/p/…
2
59
openai red teaming had two ai agents running in separate virtual machines. no network connection between them. no shared writable files. they weren’t supposed to be able to communicate.
The agents figured out how to.
one manipulated what was stored in the processor’s cache, the other measured tiny differences in memory access times to decode the message.
Fascinating Synthetic Outlaw behavior.
37
Al can follow the rules and still cause the harm those rules were meant to prevent.
That is the Synthetic Outlaw.
Our independent brief for the White House Super Intelligence Force is now available at:
syntheticoutlaw.com/whitehou…
1
54
The White House has 120 days to define AI / SI risk.
Its plan includes voluntary audits and labs policing themselves.
We've spent about 2 years researching this problem and believe that this approach is problematic.
The Synthetic Outlaw Observatory Foundation already demonstrates this from real world events.
Transformative technology needs transformative governance.
I outline it in the book:
amazon.com/dp/B0HL94VSGW
The Observatory data is here:
syntheticoutlaw.com/observat…
@WSJ @haydenfield
1
1
43
@elonmusk why would this post be throttled as political? I have been noticing that with quite a few of my posts recently.
15
Singapore just published new AI risk rules for financial institutions.
MAS recognizes the problem: AI agents can autonomously take actions that diverge from the goals humans gave them.
Its answer includes oversight, audit trails, accountability and kill switches.
But there is a deeper problem.
A kill switch solves interruption, not detection.
If an autonomous system can reach a prohibited outcome through permitted actions, when exactly do you pull it?
Regulators are beginning to tackle the responsibility gap.
They have not closed the deterrence gap. That is where the Synthetic Outlaw lives.
3
50
OpenAI just released hundreds of mathematical manuscripts devised by an unreleased model tackling open research problems.
Today it is mathematics.
Tomorrow it is medicine, aging, materials science, energy and physics.
We can compress decades of human scientific progress into weeks.
Slowing AI down is not automatically the cautious position. There is a real cost to delay in all the above.
We should manage the risks as we should be building systems capable of accelerating science and quality of life.
Humanity has spent centuries wishing for another "once in a generation mind" . We can have millions of them. We just need to govern the systems so the Synthetic Outlaw does not arise.
1
2
72
McDonald's is being sued for AI price-fixing, and the defense is:
"AI does not set prices. Franchisees do."
14,000 franchisees conforming to a single pricing engine. A portal that warns owners they "may be competitors" while tracking "pricing non-compliance."
This is the Synthetic Outlaw: the rule holds on paper and dies in the portal. Independent on the form. Coordinated in fact.
DOJ already ran this case on rental housing (RealPage). Now it's burgers. The venue changes. The failure mode doesn't.
2
31
New Mexico's AI safety bill: 30-day pre-training risk assessments. 24-hour incident reporting.
The UNM attack was a burst.
Machine speed.
A 24-hour window means the agent is done and you hear about it tomorrow. A 30-day pre-training assessment reviews the model before training, not the system as deployed.
Well-intentioned. Structurally hopeless. Human-speed paperwork can't govern machine-speed agents.
That's not safety. That's theater.
The SyntheticOutlaw.com
I wrote the book.
@PatLohmann
36





