“The ability to understand the strategic and economic impact of cyber incidents has validated it as a major systemic economic threat,” the WEF said in a report issued Friday (May 15) in conjunction with its Annual Meeting on Cybersecurity.

The WEF cited the example of a major cyberattack last year which led to a prolonged shutdown of car production in the U.K., which weakened the country’s economic growth and had an impact of around 1.9 billion pounds (about $2.5 billion) and affected thousands of organizations.

Cyber incidents are increasing insurance and compliance costs and recovery expenditures, while interrupting operations, hurting customer trust and, in some cases, threatening the solvency of businesses, particularly smaller businesses, the report added.

“At the national level, weak cyber resilience can deter foreign investment, undermine innovation ecosystems and erode competitiveness in critical industries,” said the WEF. “As economies become more digital and interconnected, cybersecurity is emerging as a foundational pillar of economic security.”

During the meeting, the report said, leaders stressed that cyber risk only becomes a sustained priority after its financial impact is clearly recognized. This is fueling demand for more robust risk quantification models and consolidated economic evidence to better guide decision-making and mobilize investment, the WEF said.

“In response, organizations are shifting away from compliance-driven approaches towards measurable resilience,” the report added. “The focus is moving to how quickly systems can recover, how much loss can be avoided, and how effectively operations can continue under stress.”

According to the WEF, this shift is reconfiguring investment priorities. Instead of expanding toolsets, leaders are focusing on high-impact capabilities, boosting visibility of critical assets and dependencies, bolstering incident response readiness and promoting rapid recovery.

The report followed one earlier this month from the International Monetary Fund (IMF), which argued that policymakers need to start treating cybersecurity as a core financial stability issue given rising AI-related cyber risks.

With attacks becoming faster, automated and more sophisticated, the existing cybersecurity measures must be expanded and sharpened, the IMF wrote on its blog.

“Policymakers should prioritize robust resilience standards, supervision focused on systemic transmission channels, and close public-private collaboration on threat intelligence and incident response,” the IMF said.

The rapidly growing threat posed by AI is spotlighted by Anthropic’s Mythos, which allows even non-experts to uncover and exploit vulnerabilities in operating systems and web browsers faster than defenders can patch and remediate them.

The AI startup agreed to meet with members of the Financial Stability Board (FSB) to discuss Anthropic’s Mythos model, the Financial Times reported Monday (May 18).

The meeting was requested by Bank of England Governor Andrew Bailey, who is also a member of the FSB, a watchdog group of finance ministry officials, central bankers and securities regulators from G20 countries, the report said.

Many FSB members have grown concerned that Mythos and AI models from other tech companies in the United States could expose weaknesses in banks’ cyber defenses, according to the report.

Anthropic said last month that Mythos had “found thousands of high-severity vulnerabilities, including some in every major operating system and web browser,” adding that the fallout “for economies, public safety and national security—could be severe,” per the report.

Only a handful of companies, most in the U.S., have gotten access to Mythos due to security concerns. This has left many organizations and regulators concerned about uneven protection levels, the report said.

Among the companies that have been granted access are Amazon, Microsoft and JPMorgan Chase, according to the report. Anthropic has agreed to keep distribution limited per a request from the White House.

The FSB is putting together a report on “sound practices” for adopting AI in the financial system, which it aims to release for consultation in June, the report said.

Other regulators are also expressing concerns. The International Monetary Fund warned this month that AI-driven cyber risk should be considered a financial stability issue because attacks can impact payment systems, confidence and liquidity at the same time.

AI is “compressing the cost and skill required to turn hacking into a scale business,” PYMNTS reported last week.

Google researchers described May 11 what they believe to be the first observed case of an AI-developed zero-day exploit tied to a planned mass exploitation campaign. It’s an event that security analysts inside and outside Google see as a sign of a larger transition toward an industrial-scale cyber threat landscape.

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.

Following the hacks, which netted the attackers a total of almost $600 million, Drift shut down and plans to relaunch after receiving stablecoins from Tether; a decentralized finance (DeFi) project called Carrot that had exposure to Drift shut down permanently; and lending protocol Aave, which was used to launder proceeds from one of the hacks, needed a rescue after investors pulled $9 billion, according to the report.

What has alarmed the industry most about these hacks is that they were likely supported by AI, the report said. While that cannot be proven, cybersecurity experts said in the report that the attacks had become so much more sophisticated, so quickly, that the hackers behind them were probably helped by AI.

Beyond that, there is the looming presence of Anthropic’s Mythos AI model, which the company has kept in limited release due to the cybersecurity risks it poses, as well as the likelihood that hackers will obtain other powerful AI models.

Cybersecurity experts said in the report that AI can help hackers find weaknesses in a blockchain protocol in days or hours, rather than months, and can give anyone the skills of an elite hacker.

Crypto firms’ responses to the threat of AI include adding software that scans devices connected to a network to detect potential threats; installing circuit breakers that pause or limit transactions above a certain threshold; and, for DeFi lenders, expanding the risk framework for collateral to include cybersecurity factors, per the report.

In an update Drift provided in April while the attack on its crypto exchange was underway, the company said: “This was a highly sophisticated operation that appears to have involved multi-week preparation and staged execution, including the use of double nonce accounts to pre-sign transactions that delayed execution.”

In the Kelp DAO hack, it was reported in April that this action highlighted the risks of interconnected systems in DeFi because the failure of one piece can threaten the entire structure.

According to a Thursday (May 14) Wall Street Journal (WSJ) report, this discovery from the security firm Califunderscores a changing cybersecurity landscape, where artificial intelligence (AI) is increasingly being used to break through even the sturdiest defenses.

The researchers say their software joins together two bugs and some techniques that could corrupt the Mac’s memory and access parts of the device that should not have been able to be accessed, the report said.

This is called a “privilege escalation exploit,” the WSJ added. If combined with other attacks, hackers could use it to take over a computer.

Michał Zalewski, a security researcher who worked at Google and who reviewed Calif’s findings, said the technique is notable because Apple has put so much work into safeguarding MacOS.

Apple, which is using frontier AI models to test and patch vulnerabilities, said it is reviewing the Calif report to validate its findings.

“Security is our top priority, and we take reports of potential vulnerabilities very seriously,” a company spokeswoman told the WSJ.

The past few months, the report added, have seen the bug-finding capabilities of AI models from companies such as Anthropic and OpenAI improve to the point that many cybersecurity experts are now warning of a “Bugmageddon.”

That’s a jokey name for an unfunny problem: AI models uncovering security vulnerabilities at unprecedented levels. That could mean a lot more work for tech teams who need to plug these holes in their defenses, to say nothing of the massive cybersecurity risk.

As PYMNTS wrote earlier this week, the issue has provided a wakeup call to the enterprise software world, which had “long held on to the traditional assumption that while bugs are common, the expertise to exploit them, particularly at scale, is rare.”

One of the things disproving this assumption is a new report from Google showing that entire attack chains are “increasingly becoming software-defined and executed faster and cheaper than ever before,” PYMNTS added, saying that the result is “not simply more hacking” but “the industrialization of hacking.”

Google Cloud researchers detailed what they say is the first observed case of an AI-developed zero-day exploit tied to a planned mass exploitation campaign.

It’s “an event that security analysts inside and outside the tech giant increasingly view less as an isolated milestone than as an early signal of a broader structural transition toward an industrial-scale cyber threat landscape,” PYMNTS added.

The system, dubbed “MDASH,” was introduced this week along with the revelation of 16 new vulnerabilities it uncovered in various versions of Windows, tech news website GeekWire reported Wednesday (May 13).

According to the report, MDASH was able to surpass Anthropic’s high-profile Mythos model on a “leading cybersecurity benchmark,” employing 100-plus specialized artificial intelligence (AI) agents working in tandem across multiple models to uncover real-world software vulnerabilities.

That metric is called the CyberGym benchmark, created by UC Berkeley researchers to determine how well AI systems can replicate real-world vulnerabilities across 1,507 tasks pulled from 188 open-source software projects. MDASH scored 88.45% on the test, with Mythos at 83.1% and OpenAI’s GPT-5.5 at 81.8%, the report said.

MDASH (“multi-model agentic scanning harness”) works by assigning different agents to do different jobs, the report added. Some scan code for potential vulnerabilities, while another group debates whether each discovery is real and exploitable. A final group puts together proof-of-concept attacks to confirm the bugs are real.

Mythos, on the other hand, is a single AI model operating inside an agent framework, GeekWire said. The startup has limited its release to a small group of companies—Microsoft included—known as “Project Glasswing.”

In the wake of Mythos’ release, OpenAI has introduced Daybreak, its own agentic security offering that works with the company’s Codex coding tool.

“AI is already good and about to get super good at cybersecurity; we’d like to start working with as many companies as possible now to help them continuously secure themselves,” OpenAI CEO Sam Altman wrote on social media platform X earlier this week.

This week also saw reports that French AI startup Mistral was working with banks in Europe—which lack access to Mythos—on its own cybersecurity offering.

In related news, PYMNTS wrote earlier this week about “the industrialization of hacking” after researchers at Google reported they had uncovered what they believe is the first observed case of an AI-created zero-day exploit tied to a planned mass exploitation campaign.

The chief takeaway for businesses is that the “tool kit of hacking tasks” for cyberscammers, including reconnaissance, exploit adaptation, vulnerability discovery and social engineering, no longer need the same level of human expertise.

“On top of that, they are all becoming increasingly automatable,” PYMNTS added. “This first-principles shift matters because cybersecurity is ultimately an economic system. And economic systems change rapidly when the cost of production collapses.”

After all, the infrastructure powering vital advances like instant payments and personalized financial services is also creating sprawling new security risks as banks connect to AI tools, FinTech solutions and third-party APIs for the thousands of financial software integrations on offer in today’s landscape. Information that once lived inside monolithic core banking systems now flows continuously across interconnected software layers designed for speed, personalization and real-time decision making.

A recent disclosure filed with the U.S. Securities and Exchange Commission (SEC) this month by U.S. commercial bank Community Bank illustrates the growing challenge of data sprawl for banks, particularly smaller and mid-size lenders looking to stand up digital innovation in order to compete with larger peers. The bank, a wholly owned subsidiary of CB Financial Services, voluntarily disclosed that an amount of sensitive customer information determined to be “material” had been exposed through an unauthorized AI application used within its environment.

The filing underscored an uncomfortable reality facing the industry: the modern banking perimeter is no longer clearly defined. The issue is not simply that banks are adopting more technology. It is that the architecture of modern banking increasingly depends on constant data mobility.

Read more: The End of the Artisanal Hack: How AI Industrialized Cybercrime 

Why Banks Are Losing Sight of Their Data

For decades, banks operated on a relatively simple security premise: protect the perimeter, secure the core and tightly control access to customer data. Sensitive information largely stayed within institution-owned systems, moving slowly through carefully managed channels and governed by rigid internal protocols. That model no longer exists.

Open banking frameworks, embedded finance partnerships and real-time payments have accelerated API adoption across the industry. Financial institutions now routinely integrate with FinTech providers for everything from fraud prevention and lending to customer onboarding and treasury management. At the same time, generative AI tools are rapidly becoming embedded inside employee workflows, customer service operations and internal analytics platforms.

Each integration creates value. Each integration also creates another potential exposure point. The challenge of defending, and even just governing, these exposure points is particularly acute for mid-sized and regional banks operating with leaner compliance and cybersecurity resources than the largest national institutions.

For example, across the credit union (CU) landscape, PYMNTS Intelligence research found that fraud now occurs across the full CU member life cycle, from account opening and onboarding to authentication and transaction activity. CUs must now defend every interaction point rather than a single stage, and 77% of CUs have experienced unauthorized network access in the past year.

The same technologies driving operational efficiency and customer personalization also increase organizational exposure. AI systems require data access to generate value. APIs require connectivity to function effectively. Modern banking infrastructure is inherently designed for openness and interoperability.

See also: The Enterprise Security Stack Is Moving to the Edge

The End of the Closed-Core Era

The real question is whether banks can establish governance models sophisticated enough to match the complexity of the ecosystems they now depend on. What has changed is the scale, speed and opacity of modern data movement. As customer data becomes increasingly distributed across external systems, governance itself is emerging as a competitive differentiator.

Rather than attempting to seal off every endpoint, many smaller institutions are shifting toward continuous monitoring models built around identity management, behavioral analytics and real-time visibility into data movement. Increasingly, the focus is less about defending a fixed perimeter and more about understanding how information flows across interconnected systems.

Data in the report “Embedding Security: Designing Fraud Risk Out of Business Transactions,” a March PYMNTS Intelligence Business Payments Tracker Series report in collaboration with WEX, reveals that nearly a quarter of banking CEOs (24%) are prioritizing AI investments for cybersecurity.

The broader banking landscape is also hoping that a rising security and data governance tide can lift all boats. PYMNTS covered Tuesday (May 12) how JPMorganChase is making nearly $14 million in philanthropic investments to support seven organizations that are combating fraud and scams through consumer awareness and real-time prevention.

Ultimately, the institutions succeeding in this transition are generally not those attempting to halt technological change. They are the ones redesigning governance around the assumption that data mobility is now permanent. Because in the API economy, the most important security question is no longer whether data leaves the bank. It is whether the bank still knows where the data went.

That was the central theme of “Embedding Security: Designing Fraud Risk Out of Business Transactions,” a March PYMNTS Intelligence Business Payments Tracker Series report in collaboration with WEX.

The report found that embedded payments are becoming core infrastructure for modern business platforms, helping companies move money inside the software they already use to run operations. That can make payments faster, easier and more visible. It also changes where fraud risk lives, spreading it across platforms, APIs, third-party partners and workflows.

The optimistic takeaway is that the same embedded structure creating new risks can also give companies more places to stop fraud earlier.

Embedded finance has moved beyond the experimental stage. WEX identifies it as one of the top business payment trends shaping 2026, with transaction value projected to exceed $7 trillion, nearly three times the $2.6 trillion recorded in 2021.

For businesses, the appeal is practical. Payments can become part of everyday work instead of a separate process. Companies can gain more control over spending, improve cash-flow visibility and reduce manual steps.

The risk is that older fraud tools were built for a slower and more centralized banking model. They often look for suspicious activity once a transaction is already in motion. In embedded payments, that may be too late.

Transactions can move through APIs and instant rails in seconds. Responsibility may be shared among banks, FinTechs, software providers and end platforms. Fraud can appear in the seams between those players, where visibility is harder to maintain.

The report pointed to several data points that show the promise and the pressure:

• Fraud attempts targeting embedded finance products are estimated to be growing two to three times faster than those across traditional banking channels. That suggests fraud is following the same growth path as embedded payments themselves.
• Fraud concerns have caused 35% of organizations to delay embedded finance and banking-as-a-service initiatives. The demand is there, but the risk is still slowing execution.
• Embedded finance is credited with reducing fraud risk by 74% of users. That finding suggests embedded payments can become safer when controls are built directly into workflows.

The shift is from detecting fraud later to limiting opportunity earlier. Virtual cards show how that can work. A company can set spend limits, merchant restrictions and authorization rules before a payment is made. Role-based permissions can determine who is allowed to initiate or approve transactions.

Multifactor authentication can add protection at access points. Real-time monitoring can flag unusual behavior as it happens rather than after money has moved.

Artificial intelligence is also becoming part of the fraud toolkit, although the report framed it as one layer in a broader system. KPMG data cited in the tracker found that 70% of banking CEOs plan to allocate 10% to 20% of their budgets to AI in the coming year. Enhanced cybersecurity is the top-reported benefit of AI adoption, cited by 24% of banking CEOs.

The broader message is that embedded payments do not have to trade speed for safety. Done well, they can make fraud prevention more precise because controls sit closer to the transaction. Identity, permissions, payment limits, monitoring and enforcement can all work together inside the flow of business. That gives companies a path to scale embedded payments with more confidence.

At PYMNTS Intelligence, we work with businesses to uncover insights that fuel intelligent, data-driven discussions on changing customer expectations, a more connected economy and the strategic shifts necessary to achieve outcomes. With rigorous research methodologies and unwavering commitment to objective quality, we offer trusted data to grow your business. As our partner, you’ll have access to our diverse team of PhDs, researchers, data analysts, number crunchers, subject matter veterans and editorial experts.

One thing the banks have learned is that Mythos can create a high-risk vulnerability by bringing together several lower risk weaknesses, according to the report.

The number of low- to moderate-ranked vulnerabilities found in banks’ tech by Mythos numbers between several hundred and thousands.

In response, banks are patching vulnerabilities in days rather than the weeks they used to take and are upgrading technology that is at the end of its software support, per the report.

Some banks may take their systems offline more often to handle the new workload but will do so in the least disruptive ways.

The testing banks are now doing with Mythos is likely to be repeated with other new AI products on a continuous basis.

For banks that don’t have access to Mythos, Anthropic has released recommendations for improving their defenses and is offering another program called Claude Security that scans for vulnerabilities and is available to a greater number of organizations, per the report.

Anthropic CEO Dario Amodei said May 5 that financial services companies and other organizations have six to 12 months to fix vulnerabilities in their software before Chinese AI models develop capabilities equal to those of Mythos.

Amodei said that Mythos had uncovered tens of thousands of vulnerabilities and that if code is rewritten with models like Mythos, “we could be in a better position than we started in because we fixed all these bugs.”

On May 3, Treasury Secretary Scott Bessent said that American banks are working to safeguard against AI-related cyberthreats and that he told them that they should take the Mythos model seriously and use it to find holes in their defenses.

“What we’ve had in the past month was a step change in the power of one large language model, but we’re going to see it from the other AI companies, and it’s important that the U.S. stays ahead here,” Bessent said.

That scarcity has helped keep digital risk manageable, if never fully contained.

But a new report from Google published Monday (May 11) is turning that legacy assumption on its head by showing where artificial intelligence is compressing the cost and skill required to turn hacking into a scale business. The report revealed that entire attack chains are increasingly becoming software-defined and executed faster and cheaper than ever before.

The result is not simply more hacking. It is the industrialization of hacking.

Google Cloud researchers described what they believe to be the first observed case of an AI-developed zero-day exploit tied to a planned mass exploitation campaign, an event that security analysts inside and outside the tech giant increasingly view less as an isolated milestone than as an early signal of a broader structural transition toward an industrial-scale cyber threat landscape.

The key takeaway for enterprise leaders and chief financial officers assessing their firms’ new risk profiles is that nearly the whole tool kit of hacking tasks for fraudsters, including reconnaissance, exploit adaptation, vulnerability discovery, social engineering and more, no longer requires the same degree of specialized human expertise. On top of that, they are all becoming increasingly automatable. This first-principles shift matters because cybersecurity is ultimately an economic system.

And economic systems change rapidly when the cost of production collapses.

See also: The Enterprise Security Stack Is Moving to the Edge

Cyber Attacks Shift From Craft Production to Mass Manufacturing

In practical terms, the ability for adversarial cyber threat groups to benefit from software-like scale efficiencies means enterprises are confronting a future in which sophisticated attacks are no longer exceptional events. They are becoming operationally routine.

When the marginal cost of generating attacks falls, the volume of attacks rises. The software industry has seen this dynamic repeatedly. Cloud computing reduced infrastructure costs and enabled startup proliferation, while social media collapsed publishing barriers and flooded the information ecosystem with content. Generative AI is now applying the same logic to cyber operations.

While traditional cyberattacks resembled artisanal production, AI has changed the production function for cybercrime. Whereas the compute skill and cost economics once limited high-end offensive capability to a relatively small set of fraudsters, the emerging danger is that AI enables ordinary attackers to operate with previously unattainable efficiency.

A phishing email no longer needs to be brilliant if millions can be generated and adapted instantly for different industries, executives and geographies. Malware no longer needs to be elegantly engineered if AI-assisted iteration allows attackers to rapidly test variations against defenses.

This can ultimately lead to attack surface saturation, where enterprises face a continuous stream of low-cost, semi-customized intrusion attempts generated at machine speed. In this environment, the sheer volume of threats becomes strategically significant even if individual attacks remain imperfect.

Read also: Cybersecurity’s Hottest New Job Is Negotiating With Hackers

How Firms Are Navigating Today’s Industrial-Scale Threat Landscape

Experienced attackers can still outperform automated systems in complex intrusions, but the threshold for good-enough offensive capability is rapidly dropping as the economics begin to favor persistence at scale.

For example, across the credit union (CU) landscape, PYMNTS Intelligence research found that fraud now occurs across the full CU member life cycle, from account opening and onboarding to authentication and transaction activity. CUs must now defend every interaction point rather than a single stage, and 77% of CUs have experienced unauthorized network access in the past year.

That does not mean defenders are powerless. AI also offers defensive advantages in detection, anomaly analysis, incident response and threat intelligence. Research from the PYMNTS Intelligence report “The AI MonitorEdge Report: COOs Leverage GenAI to Reduce Data Security Losses” showed that 55% of companies are employing AI-powered cybersecurity measures.

Still, in today’s threat environment, cybersecurity strategies increasingly resemble industrial risk management rather than perimeter defense. Few sectors may feel this shift more acutely than cyber insurance, where the existing market was built on actuarial assumptions pricing risk around relatively observable controls, including endpoint security, employee training, patch management, multifactor authentication and incident response maturity.

These assumptions look increasingly unstable in an AI-driven threat environment. Firms previously considered moderate risks could suddenly face elevated exposure simply because attackers can now economically target a much broader universe of companies.

As AI lowers barriers for attackers, the standard for what constitutes reasonable defense may evolve upward as well. Companies with mature security architectures, strong identity controls, segmented infrastructure and rapid patching capabilities may increasingly resemble low-risk operators in a high-risk economy.

The central executive question is no longer whether a company can prevent every intrusion. It is becoming whether the organization can remain operationally resilient in a world where sophisticated attacks become a continuous background condition of doing business.

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.

Writing in the latest GTIG AI Threat Tracker, which was released Monday (May 11), GTIG said a criminal threat actor planned to use the zero-day exploit in a mass exploitation event, but GTIG may have prevented it with proactive counter discovery.

After identifying the zero-day vulnerability in a Python script that enables the user to bypass two-factor authentication on an open-source, web-based system administration tool, GTIG worked with the impacted vendor to responsibly disclose the vulnerability and disrupt the threat activity, according to the report.

GTIG said it has “high confidence” that the threat actor used an AI model to discover and weaponize the vulnerability.

“As the coding capabilities of AI models advance, we continue to observe adversaries increasingly leverage these tools as expert-level force multipliers for vulnerability research and exploit development, including for zero-day vulnerabilities,” GTIG said in the report. “While these tools empower defensive research, they also lower the barrier for adversaries to reverse-engineer applications and develop sophisticated, AI-generated exploits.”

Other AI-related threat activity highlighted by GTIC in the report includes AI-augmented development for defense evasion, autonomous malware operations in which models dynamically generate commands, and AI-augmented research and information operations campaigns that generate synthetic media and deepfake content at scale.

The report also spotlighted obfuscated LLM access, in which threat actors pursue anonymized access to models to illicitly bypass usage limits, and supply chain attacks in which adversaries target AI environments and software dependencies as an initial access vector.

“Attackers rarely shy away from experimentation and innovation, but neither do we,” GTIG said in the report. “In addition to sharing our findings and mitigations with the larger security and AI community, Google employs proactive measures to stay ahead of these constantly changing threats.”

In earlier editions of the GTIG AI Threat Tracker, the organization noted a new form of intellectual property theft called “model extraction attacks” or “distillation attacks” and threat actors’ use of AI for not only productivity gains but also “novel AI-enabled operations.”

The International Monetary Fund (IMF) said in a Thursday (May 7) blog post that at a time of rapidly accelerating cyber risk driven by AI, cybersecurity is a core financial stability issue and should be treated as such by policymakers.