SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios.
The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios.
Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said.
Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior.
Read my latest for Axios here: axios.com/2026/10/22/ai-…
The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios.
Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said.
Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior.
Read my latest for Axios here: axios.com/2026/10/22/ai-…
Oct 22, 2026 · 12:48 PM UTC
1,360
6,880
19,047
8,913,765
@MadisonMills22 Start suing and prosecuting the AI companies.
37
162
2,392
122,160
@MadisonMills22 I’m calling my wife’s spending “rogue agent behavior”
8
13
504
51,247
@MadisonMills22ive in a corrupt oligarchy increasingly controlled by megalomaniacal techno-fascists and an increasingly demented criminal president we might have been able to use AI productively.
As it stands there is almost no chance any of this ends well. This is just the tip of the chaos iceberg.
As it stands there is almost no chance any of this ends well. This is just the tip of the chaos iceberg.
25
75
450
24,197
@MadisonMills22 Whatever agent that’s in my account can’t be more financially damaging than my teenage daughter
64
25
2,441
178,711
@MadisonMills22 We’ve had almost 7 months to prepare and over 3 years to know these type of attacks are coming. It’s seriously pathetic there hasn’t been safeguards or a framework implemented years ago.
55
58
1,296
102,603
@MadisonMills22 This is an extraordinary claim, even given the seriousness of the php2shell vulnerabilities.
Is there any independent verification that this is happening?
Please share links to any other serious reporters covering this, independent of Axios.
Is there any independent verification that this is happening?
Please share links to any other serious reporters covering this, independent of Axios.
7
5
122
13,871
@MadisonMills22 FAKE
These companies are committing CRIMES stop calling it the agents. It's never the agents.
These companies are committing CRIMES stop calling it the agents. It's never the agents.
10
11
239
16,265
🚨 ¿REALMENTE ENTIENDES LO QUE ESTA PASANDO AHORA MISMO?
Todo el mundo temia que la IA acabara con crypto
Pero resulta que quiza el primer objetivo sea tu banco
Segun Axios, JPMorgan, Bank of America, Wells Fargo, Morgan Stanley y Citi estan detectando agentes de IA dentro de ntes
Entraban con los logins que los mismos usuarios les daban
Algunos llamaban a los call centers
Otros incluso parecian actuar por su cuenta
Los bancos ahora mismo no pueden distinguir bien entre un humano y un agente
Ni saber si el cliente autorizo cada movimiento o si el agente se salio de control
Morgan Stanley ya los esta bloqueando
JPMorgan, Bank of America y Citi estan restringiendo el acceso
Y ni siquiera hizo falta hackear nada
La gente le entrega a un agente su usuario y contraseña y lo deja manejar su dinero
Los bancos ya se preparan para el peor escenario: como asegurar o restaurar cuentas si algo sale mal
Solo que esto no es un hack tradicional
Asi que si la persona dio acceso a su cuenta, esas transferencias pueden no contar como no autorizadas
O sea, puede que nadie se las reembolse
Y los bancos siempre vendieron lo contrario
Si alguien te roba, ellos responden
Ahora la pregunta es si eso sigue aplicando cuando el alguien es una IA
1
3
14
1,799
@MadisonMills22 Hold A.I. companies and executives responsible for crimes committed by their agents.
13
40
580
51,250
@MadisonMills22 they want to take away your ability to automate your finances
11
7
225
25,308
@MadisonMills22 Yeah - we'll look back at this in a couple months and... probably not even remember it because of all the other even more concerning things which will happen between now and then.
12
13
303
46,542
@MadisonMills22 Fk the banks and axios. They will be behind the financial crash and blame it on “rogue AI” because they want a reason to scare people into accepting their bs agenda to maintain financial control. But it won’t work. Too many people are awake !
0
7
47
6,191
@MadisonMills22 I’m starting to think this is why USAA was offline this weekend. Customers had no access to debit transactions for almost eight hours with no communication from the bank about what was happening, or why members couldn’t use their cards at the grocery store or gas station.
8
41
233
22,177
@MadisonMills22 These agents aren't breaking into random people's accounts FYI. People who are using them are allowing the agents access to their own accounts and the agents are doing more than they're supposed to.
7
3
45
6,066
@MadisonMills22 These are all things that I have been teaching since 2016 to my high school students. We as a society are not anywhere close to understanding AI and able to create policy for it.
Teaching AI for 10 years has taught me alot about our failures in policy.
Teaching AI for 10 years has taught me alot about our failures in policy.
2
19
87
11,391
@MadisonMills22 一个可参考的协议工具是 OAuth RAR（RFC 9396）：它的示例将读账户与发起支付分开，支付授权可写明金额、币种和收款方，由授权服务与银行 API 共同执行限制。这样，“让代理看看账单”就不必等于允许它任意转账。具体银行是否支持，仍需逐家核对。 rfc-editor.org/rfc/rfc9396.ht…
0
0
1
1,718
@MadisonMills22 If a customer gave access to their investments and accounts to an AI Agent should have no way to complaint if anything goes wrong.
Seriously 🙄🤷♂️
Seriously 🙄🤷♂️
1
3
12
3,221
@MadisonMills22 What type of moron gives an AI agent their banking credentials? Darwin Award finalists.
3
2
39
3,168
@MadisonMills22 Literally false. The report itself notes that the AIs were authorized by the account holders to access their own accounts. The “unauthorized” part refers to the bank claiming this usage violates some ToS. But this is by no means illegal hacking.
1
0
7
2,120
@MadisonMills22t but this is probably the first time ever where it feels **safer** to be broke. You’re not sitting here stressing about not if, but when AI will wipe out your millions in the bank, home values, businesses…
You’re just waiting for 💩 to hit the fan, and see the elites freaking out😂
You’re just waiting for 💩 to hit the fan, and see the elites freaking out😂
2
10
49
2,441
@MadisonMills22 Y'all are acting like rogue AI is a surprise when banks have had weak security for years. maybe I'm dumb but how does an unauthorized AI bot even log into a customer account
5
3
49
5,389
@MadisonMills22 I think they’re more alarmed about how many people are either canceling services, moving accounts, and/or taking any action that favors the customer and saves them money.
Remember, that “cancel” option is buried very deep on purpose!
Remember, that “cancel” option is buried very deep on purpose!
1
1
23
6,845
@MadisonMills22 “I promise! It was just some AI bot logging in and subscribing to OnlyFans.”
0
0
8
1,100
@MadisonMills22 "AI agents" is code word for criminal malware. AI companies are really hacking collectives producing malicious computer viruses to attack computers.
Unleashing malware has been a crime in America for many years.
Unleashing malware has been a crime in America for many years.
0
2
6
102
@MadisonMills22 I suggest at a minimum, go back to paper statements and paper records. At least those survive an AI attack
0
1
2
242
@MadisonMills22 could you be more vague?
major banks, banking executives, taking actions,
what is happening and where?
major banks, banking executives, taking actions,
what is happening and where?
4
0
39
9,490
@MadisonMills22 This is all just as ridiculous and annoying as it was predictable.
0
2
39
14,789
@MadisonMills22 Republicans: let AI companies regulate themselves
0
1
6
1,880
@MadisonMills22 This woman is sensualizing this. She even put in the damn article they where giving access by the account holders!
0
2
9
1,828
@MadisonMills22 @BretBaier So basically, they know nothing. They are panicking and have no idea why but feel like they should be.
0
0
3
1,055
@MadisonMills22 Is this some of the bad stuff we are supposed to just accept for their greater good?
0
2
22
5,088