Skip to content

What Is Happening to US Banks? JPMorgan Disruption and Growing Cybersecurity Threats Explained

Shalu Shrotiya, Digital Marketing Manager 20 min read

Share

JPMorgan Cybersecurity After the Outage

By mid morning on Monday, November 16, customers of Chase, Wells Fargo, Bank of America and several other US lenders were staring at login screens that would not load. Downdetector reports for Chase passed 4,000 by midday Central time and 5,000 by about 2 p.m. Eastern, depending on which snapshot you read. Within the hour, social media had reached a verdict: cyberattack.

JPMorgan Chase said otherwise. A spokesperson described a software bug in on premises hardware systems and said there was no evidence the problem was cyber related. The bank added that it was not a systemic issue, and that other banks were affected only where they were a party to a transaction with Chase.

Take the bank at its word for now, because nothing public contradicts it. But notice how fast the public assumed the worst. In a year shaped by a US and Iran war, record cybercrime losses and a Google report claiming that autonomous attacks have arrived, a bank outage no longer reads as an IT story first. It reads as a security story until someone proves otherwise. For anyone responsible for US banking cybersecurity, or for the security of any enterprise that leans on the same payment rails, cloud platforms and identity providers, that reflex is worth understanding.

This article separates what is confirmed from what is claimed, explains the real threat picture behind the "cyber world war" headlines, and sets out what IT leaders can do before the next disruption, whatever its cause.

US Banking Cybersecurity: Key Lessons from the JPMorgan Disruption

The November 16 disruption was, by the bank's account, a software failure and not an attack. The more important lesson is that a failure inside one large institution spread to others within hours, and that the threat environment around it is genuinely worse than a year ago. State backed groups remain active, attackers now have AI tools that can turn a published patch into a working exploit, and the rules on incident reporting are in flux. Enterprises that map their dependencies by business service, shorten their patch windows, rehearse the "is this an attack?" question and line up their reporting clocks will cope with both accidents and attacks. Those that treat the two as separate problems will lose time on the day it matters.

What actually happened on November 16, and what nobody has proven

The confirmed facts are narrow. Chase attributed the disruption to a software bug in on premises hardware systems and said customers retained access to their accounts and funds. Wells Fargo confirmed online banking problems while saying ATMs and branches stayed available. Zelle said transfer problems stemmed from a Chase issue, and reports at other banks appeared tied to sending money to Chase accounts. U.S. Bank said its own systems were not directly affected. Plaid, which connects bank accounts to third party apps, flagged disruption to its Chase connections on its status page.

According to Downdetector figures reported that day, about a third of complaints about Chase concerned the website, 32% involved fund transfers and 20% involved logins. That pattern fits a failure in shared transaction and authentication plumbing more than it fits a campaign against several separate banks.

Context matters here, and so does restraint. Four days earlier, on November 12, Microsoft 365 suffered a long outage that disrupted Exchange Online, Teams, SharePoint and Copilot. Microsoft said it saw no indication of a cyberattack and described failed server health checks, though a full post incident report had not appeared in the coverage reviewed. On November 13 the FAA issued a ground stop, and the agency said its investigation had found no sign of a cyberattack. United Airlines also reported a technology outage on the 16th. No agency and no affected company has linked any of these events to one another or to a foreign actor.

That is the right posture for a technology leader too. A cluster of outages is a reason to ask questions. It is not evidence of coordination.

The claim that should be kept separate: the n0n extortion report

On the same day, threat intelligence firm Hackmanac flagged a claim that an extortion group called n0n had stolen customer and marketing data from Visa, Mastercard and JPMorgan Chase, with a threat to publish unless a settlement was reached. As of the coverage reviewed, the claim was unverified, none of the three companies had issued a statement, and the date reflects when the claim was observed, not when any intrusion occurred. Extortion groups routinely exaggerate. It may still turn out to be nothing. It has no demonstrated connection to the outage, and the two should not be merged in a board paper or a LinkedIn post.

Is a "cyber world war" really underway?

The backdrop is the war that began on February 28, 2026, when the US and Israel struck Iran. Within days, Reuters reported that US financial firms were on heightened alert, and regulators in New York and California warned lenders that recent events called for vigilance. NYDFS urged regulated firms to review their cybersecurity programs against a heightened threat environment. FINRA's alert, as of March 16, said it knew of no significant Iran related attacks on financial services but was aware of reporting that Iranian actors were actively targeting US financial institutions and critical infrastructure.

A ceasefire was announced in April, and security researchers and AP reporting noted almost immediately that it did not extend to cyber operations. An Iran linked group called Handala said publicly that its cyber campaign would outlast any military pause. In June, US officials said the Iranian cyber threat would persist even as a preliminary diplomatic deal took shape. I could not confirm the current state of the conflict itself, so this article does not characterize it.

Iran is not the only actor. The Office of the Director of National Intelligence's 2026 Annual Threat Assessment, as summarized by security media, describes China as the most active and persistent cyber threat to US government and private sector networks, with investment aimed at pre positioning for disruptive attacks. It describes Russia as a persistent and advanced threat, and Iran and North Korea as less capable but increasingly willing to use cyber operations against the US.

At the criminal and hacktivist layer, Intel 471's 2026 financial sector report tracked 562 DDoS attacks claimed against financial targets in its reporting window, led by NoName057(16) and Dark Storm Team, and noted that the Everest extortion operation claimed two US banks in April 2026. It also noted a $1.5 billion crypto heist attributed to a nation state group.

My reading, which is analysis and not a finding: "attrition" is a better word than "war." Attackers do not need a single decisive blow against a bank. They need enough friction, across enough targets, that trust and operations erode. A software bug that looks like an attack achieves some of that effect without any attacker at all.

Why did November 2026 become the week machine speed attacks went on the record?

On November 16, Google Threat Intelligence Group and Mandiant published their Q4 AI Threat Tracker under a blunt headline: the era of autonomous attacks has arrived. The centerpiece is a framework GTIG calls exploit_agent, found on an exposed staging server. According to the report, GTIG identified threat actors using LLM frameworks to fully automate vulnerability discovery and exploitation without a human in the loop for the first time. The framework watches for software patches, compares code before and after, works out what was fixed, drafts an exploit and runs it. Its observed targets were French websites in media, entertainment and education. Google did not name an operator, though comments in the configuration were in Russian.

Three details matter for banks. First, these tools target flaws that already have a fix, which means the gap between a vendor patch and your deployment is now the main exposure. Second, GTIG reports that zero day exploitation averaged eight a month in 2025, rose to 11 a month in 2026 and spiked to 22 in August. Third, Google's report projects simple attacks with little human involvement within one to three months and sophisticated, fully autonomous attacks in three to six months. That last item is a forecast from a vendor that also sells defensive AI products, so treat it as informed but interested.

The FBI's own warning came three days earlier. In a flash alert on November 13, the Bureau said criminals and other actors have been using ARTEX, an open source, LLM driven agentic platform, since at least August to automate intrusion stages "at machine speed." It added that some intrusions moved from initial targeting to privileged access in under three hours. Microsoft threat reporting separately indicates that a financially motivated actor tracked as Storm-3358 began targeting financial sector organizations with ARTEX in October.

One caution from practitioners is worth repeating. Names like ARTEX will change. Attackers can swap tools and models next week, so detection built around a framework's name ages quickly. Detection built around behavior, such as unusual enumeration, credential abuse and rapid privilege escalation, does not.

What do the numbers say about the cost of being a bank in 2026?

Several credible datasets point in the same direction, with different scopes. It helps to line them up.

Breach cost. IBM's Cost of a Data Breach Report 2026 puts the average financial services breach at $6.29 million, against $5.56 million a year earlier and a global cross industry average of $4.99 million. Healthcare remains highest at $6.64 million, though it fell about 10%. Secondary write ups describe the financial services rise as 12% or 13%. The arithmetic on the two published figures gives roughly 13%, and the difference appears to be rounding or method. Either way, finance is the second most expensive sector and moving the wrong way.

For Indian readers. IBM's India edition reports a record average breach cost of INR 25.5 crore, with financial services highest at INR 40.9 crore. Shadow AI, regulatory non compliance and cloud migration were the top cost drivers, with shadow AI alone adding INR 1.79 crore on average.

Bank executive sentiment. KPMG's 2026 Banking Technology Survey of 200 US banking executives found 76% had seen an increase in attacks over the past year and 92% were increasing cyber budgets. In the same survey, 69% named cloud providers as their biggest concentration concern.

Large enterprise view. KPMG's separate 2026 Cybersecurity and Technology Risk Survey of 310 security leaders at US organizations with more than $1 billion in revenue found 83% reported more attacks, but only 24% had fully integrated AI into their security programs, while 53% had partial adoption.

Crime losses. The FBI's 2025 Internet Crime Report recorded $20.877 billion in reported losses, up 26%, across 1,008,597 complaints, the first time the million mark was crossed. Crypto related losses were $11.366 billion. For the first time the report carried a section on AI, citing 22,364 complaints and roughly $893 million in losses. The FBI notes that figures rely on victim reporting and likely undercount.

What do these numbers mean together? Budgets are rising almost everywhere, yet adoption of the tools that match attacker speed is uneven. Spending is not the gap. Integration and speed are.

What does the outage reveal about concentration risk?

Set the attack question aside and look at what the November 16 event shows by itself. One institution had a fault. Customers of other banks then reported problems because they were sending money into it. A data aggregator flagged trouble. A payment network said the issue sat with the bank. Nobody had to be hacked for the effect to spread.

That is concentration risk made visible. It also shows up in the survey data above, and in older cases. The 2024 attack on C-Edge Technologies in India forced nearly 300 small banks offline because they shared one technology provider. The Microsoft 365 outage of November 12 and 13 disrupted organizations across dozens of countries, with at least one report describing firms falling back to fax machines to confirm trades.

A new layer is arriving as well. In JPMorgan's most recent earnings call, CEO Jamie Dimon said of outside AI agents that "you don't want rogue agents," signaling the bank will not open customer accounts to just any third party agent. Whatever one thinks of that stance, it shows where a new dependency is forming: software acting on a customer's behalf, with credentials, at machine speed. Every dependency is also a blast radius.

Which US cyber rules are changing, and why does it matter outside the US?

The US rulebook is moving in two directions at once. Some changes would add reporting duties, others would remove or soften them, and several have no settled date. Global enterprises and Indian firms with US operations should follow all four.

SEC incident disclosure. The 2023 rule requires public companies, including listed banks, to disclose a material cyber incident within four business days of determining it is material. Five major trade groups, including the ABA, argue it hands attackers a roadmap and have pressed the SEC to rescind it, most recently in April 2026 comment letters, while SEC Chair Paul Atkins reviews corporate disclosure rules more broadly. A Debevoise tracker cited by American Banker counted 29 material incident filings in the two years after the SEC clarified its scope in 2024, versus 50 companies using a voluntary route. No rescission had been finalized in the coverage reviewed.

CISA 2015 information sharing protections. The law that shields companies from liability when sharing threat data with the government lapsed for roughly six weeks in autumn 2025, was extended several times and now expires on December 11, 2026. A coalition of 35 groups, including the ABA, wrote on November 1 asking Congress to extend it to 2035 through the defense authorization bill. A dispute in the Senate over an unrelated speech provision has held up a longer deal. For banks, this is a live deadline in less than four weeks.

CIRCIA. CISA sent its final incident reporting rule to the White House Office of Management and Budget on October 1, 2026, after missing the original October 2025 deadline and two later targets. As proposed, covered entities would report substantial incidents within 72 hours and ransom payments within 24 hours. The review can take up to 90 days, and an effective date would follow publication by at least 60 days, so the realistic start sits in 2027. Financial services is one of the 16 covered sectors, though the final scope may differ from the 2024 proposal.

NYDFS. On May 21, 2026, the New York regulator issued two industry letters, one on heightened cybersecurity risk and one on frontier AI models. Neither creates new Part 500 rules. Law firm analyses say both signal what examiners will ask, including refreshed risk assessments that account for frontier AI threats. A further guidance on risk assessments followed on September 10.

The practical effect for IT leaders is that one serious incident could start several clocks together: an SEC materiality decision, a state regulator notice, a CIRCIA report, contractual notices to partners and, for global firms, local rules such as RBI requirements. Teams that run these through separate spreadsheets will lose hours.

What should enterprise IT leaders do now?

Five moves matter most, in rough order of payoff.

1. Rehearse the "is this an attack?" question. On November 16, the world guessed within an hour. Your executives and customers will too. Build a first hour decision path: who confirms whether a failure is internal, who owns external statements, and what evidence lets you say "no evidence of a cyber incident" without overreaching. Chase's statement was careful; it said "no evidence," which is different from "not an attack."

2. Measure your real patch window. Calculate the time between a vendor fix and deployment on every internet facing system, not the policy number. If machine speed exploitation of already patched flaws is now documented, a monthly change window is an invitation. Prioritize flaws listed as actively exploited and automate where change control allows.

3. Map dependencies by business service. A vendor list does not tell you what breaks. List your five or six most critical services, such as payments, customer login and treasury, and trace each through the cloud, identity, network and third party providers it needs. Then test what happens when each one fails, with and without an attacker.

4. Detect behavior and shorten containment. If intrusions can reach privileged access in under three hours, detection that depends on next day analyst review is too slow. Invest in identity monitoring, privileged access controls, network segmentation and continuous monitoring that alerts on behavior.

5. Unify the reporting clocks. Keep one incident record that feeds every obligation. Pre agree materiality criteria, legal ownership and templates. Run a tabletop where three regulators and two partners must be notified inside 72 hours.

A sixth item deserves a place on the agenda for 2027: governing AI agents, whether your own, your vendors' or your customers'. Shadow AI added INR 1.79 crore to the average Indian breach in IBM's data, so this is no longer a theoretical risk.

What happens next? A 2 to 5 year outlook

The following are forecasts and informed analysis, not confirmed facts.

Patch windows will compress further. If Google's projection is even partly right, the safe interval between a patch release and exploitation will shrink toward hours for popular software. Expect more automated patching and more pressure on vendors to ship fixes that are quick to deploy.

Resilience will rival prevention in supervisory attention. Regulators already treat cyber as a systemic risk. After a year of high profile outages, expect closer scrutiny of operational resilience, third party concentration and the ability to recover critical services, not only breach prevention.

Disclosure rules will stay unsettled. The SEC rule, CIRCIA and CISA 2015 are all moving, and the final shape is uncertain. Planning for the strictest realistic combination, rather than waiting, is the safer path.

AI will cut both ways. The same technology that speeds attacks can speed defense. IBM's 2025 report found organizations using AI and automation extensively paid about $3.62 million per breach versus $5.52 million for those without. The opportunity is real, but only for teams that integrate it into operations, which, per KPMG, fewer than a quarter have done.

Agent to agent finance is the next dependency. Customer side AI agents will request account access, move money and negotiate on behalf of people and companies. The identity, liability and fraud rules for that world have yet to be written.

The NS3TechSolutions perspective

NS3TechSolutions works with enterprises on IT infrastructure, enterprise networking, cybersecurity, cloud and managed services, including security operations (SOC) and network operations (NOC) support. The events above map onto that work in specific ways, though not every organization needs every service.

Dependency and resilience reviews, which start from critical business services and trace them through the network, cloud and provider layers, are where most organizations find surprises. Network security and segmentation work, across firewall and OEM platforms such as Fortinet, Cisco and Palo Alto Networks, limits how far an intruder can move once inside, which matters when privileged access can arrive in hours. Round the clock monitoring through a SOC and NOC shortens the time between an anomaly and a decision, and it provides the evidence trail that a regulatory clock demands. Incident response planning and tabletop exercises tie these together, particularly for Indian banks, NBFCs and enterprises that must satisfy RBI and CERT In expectations while also dealing with global customers and partners.

The approach is practical: start with what the business cannot afford to lose, test it against both accident and attack, and close the gaps in order of consequence.

Practical checklist: the first hour, the first week, the first quarter

Save this, share it with your team, or paste it into your next risk review.

First hour of any disruption

  1. Name an incident commander and an external spokesperson.

  2. Check whether the failure is internal, a provider issue or hostile activity, and record the evidence.

  3. Verify identity systems, privileged accounts and logging before assuming the cause.

  4. Use careful public language, such as "no evidence of a cyber incident" with a commitment to update.

  5. Start one incident record that can feed every regulator and partner notice.

First week 6. Calculate your true patch window for every internet facing system. 7. List all systems flagged as actively exploited and set a hard deadline for each. 8. Identify your five most critical business services and who owns each. 9. Review MFA, privileged access and service account hygiene.

First quarter 10. Map each critical service to its cloud, network, identity and third party dependencies. 11. Run a tabletop with a simultaneous outage and suspected breach, with at least two regulatory clocks. 12. Refresh your risk assessment for AI enabled attackers and AI agents, as NYDFS expects of its regulated firms. 13. Review the status of CISA 2015, CIRCIA and the SEC rule against your legal obligations each quarter. 14. Decide where managed detection and response or a SOC can close staffing and coverage gaps.

FAQ

Q. Was the November 16 JPMorgan Chase outage a cyberattack?

A. JPMorgan Chase said a software bug in on premises hardware systems caused the disruption and that there was no evidence it was cyber related. As of the coverage reviewed on November 17, no bank, regulator or agency had attributed it to an attack. "No evidence" is not proof of absence, so watch for the bank's post incident explanation.

Q. Why were other banks affected if the bug was at Chase?

A. Chase said other banks saw slowness only where they were party to a transaction with it. Zelle said the issue sat with Chase, and reports at other banks appeared tied to payments sent to Chase accounts. Interconnected payment systems let one institution's failure appear as many.

Q. Are US banks under cyberattack from Iran right now?

A. US financial regulators and industry groups have warned about elevated Iranian activity since the war began on February 28, 2026, and US officials said in June that the cyber threat persisted despite diplomacy. FINRA's March alert said it knew of no significant Iran related attacks on financial services at that time. There is no public evidence tying Iran to the November 16 outage.

Q. What is an autonomous AI cyberattack, and has one happened?

A. Google's Threat Intelligence Group reported on November 16 that it found a framework that automatically discovers, builds and runs exploits for patched vulnerabilities without human involvement. It is the first such documented case according to Google, and its observed targets were French websites. The finding is from a vendor that sells defensive AI tools, so read it with that in mind, but the FBI's separate November 13 alert on ARTEX points the same way.

Q. How much does a data breach cost a financial institution in 2026?

A. IBM's 2026 report puts the average at $6.29 million for financial services globally, up from $5.56 million, against a $4.99 million all industry average. In India, IBM reports INR 40.9 crore for financial services.

Q. Do banks have to report cyber incidents publicly, and is that changing?

A. Listed US companies must disclose material incidents within four business days of a materiality determination under the SEC rule, though trade groups are pressing for repeal. CIRCIA would add 72 hour incident and 24 hour ransom payment reporting once its final rule takes effect, which is not expected before 2027. Banks also report to their regulators under existing rules.

Q. What should Indian banks and enterprises learn from this?

A. Three lessons carry over: map dependencies by business service, because shared providers can take down many institutions at once, as C-Edge showed in 2024; shorten patch and containment times; and rehearse communications, because the first public question is always whether it was an attack.

Q. How can an IT team tell an outage from an attack in the first hour?

A. Check identity and authentication logs for unusual activity, look for signs of data movement or privilege changes, confirm whether the failure matches a recent change or hardware fault, and review provider status pages. Assign one person to this question and avoid guessing in public until evidence supports a statement.