NEW YORK, May 12 —
- AI Security Pressure: Anthropic’s Mythos AI is reportedly helping major U.S. banks uncover hundreds to thousands of software vulnerabilities.
- Bank Response: Large lenders are rushing to patch weaknesses, in some cases fixing issues in days instead of weeks.
- Smaller Institutions: Regional and smaller banks without Mythos access are relying on shared intelligence from larger institutions.
- Operational Risk: Experts warn accelerated AI-driven testing may lead to more frequent system downtime as fixes are deployed.
- Pricing Barrier: Mythos reportedly costs $25 per million input tokens and $125 per million output tokens, limiting accessibility.
- Cybersecurity Shift: Industry experts say cyber threats are moving at “machine speed”, forcing banks to rethink legacy defenses.
| Metric | Value | Context |
|---|---|---|
| Mythos input token cost | $25 per million | AI pricing model |
| Mythos output token cost | $125 per million | AI pricing model |
| Relative cost vs Opus 4.7 | 5x higher | Anthropic comparison |
| Organizations with access | ~40+ | Mythos preview participants |
| Anthropic customer credits | $100 million | Research preview support |
| Vulnerabilities found | Hundreds to thousands | Reported bank findings |
Banks Confront a New Era of AI-Powered Cybersecurity
U.S. banks are rapidly strengthening cybersecurity defenses after Anthropic’s Mythos AI reportedly identified large numbers of weaknesses across financial institutions’ technology systems.
According to sources familiar with the matter, some of the country’s largest lenders have gained access to the advanced AI tool and are now accelerating software updates, internal testing, and vulnerability remediation efforts.
The technology appears especially effective at identifying combinations of smaller software flaws that, when linked together, create larger and potentially more dangerous security risks.
For banks operating on aging infrastructure, the findings are triggering renewed urgency around technology modernization.
Why Mythos Is Causing Alarm
Cybersecurity experts say the significance of Mythos lies in its ability to detect threats at unprecedented speed.
Traditionally, vulnerabilities in banking systems could remain hidden for extended periods before attackers identified and exploited them. AI-driven systems are changing that equation.
Experts warn that machine-powered cybersecurity tools are rapidly shrinking the time institutions have to respond, forcing banks to patch systems far faster than traditional timelines allowed.
The challenge is particularly acute for financial institutions dependent on legacy software nearing the end of technical support cycles.
Industry observers say banks are now confronting a difficult balance between strengthening security and maintaining uninterrupted customer access to critical services.
Pressure on Legacy Systems
One major concern is operational disruption.
Sources familiar with banks’ internal response efforts said institutions may need to temporarily take systems offline more frequently as vulnerabilities are repaired and software upgrades deployed.
While banks are expected to minimize customer impact, increased maintenance activity could raise the risk of temporary service interruptions.
The banking sector’s dependence on complex digital infrastructure makes cybersecurity failures especially consequential, given that payment systems, deposits, lending operations, and financial markets rely heavily on uninterrupted technology.
Smaller Banks Face High Barriers
Access to Mythos remains limited.
The AI system is reportedly available only to select organizations, including some of the country’s largest banks and cybersecurity partners. Smaller institutions face barriers related to cost and computing requirements.
As a result, some regional and community banks are relying on information shared by larger institutions to anticipate emerging vulnerabilities and strengthen defenses.
That dynamic may widen the cybersecurity gap between major Wall Street institutions and smaller lenders with fewer technical resources.
Cost and Competition in AI Security
Mythos also reflects the growing commercialization of advanced AI security tools.
Anthropic prices the model based on usage, charging significantly more than some of its broader commercial AI offerings. The company has said pricing reflects the balance between accessibility and funding continued development.
At the same time, Anthropic has announced support programs and security guidance aimed at helping organizations improve cyber resilience even without direct access to Mythos.
A Banking Industry Wake-Up Call
The emergence of powerful AI vulnerability scanners signals a broader transformation in cybersecurity.
For banks, the message is increasingly clear: institutions built around older systems may need to modernize faster than expected as cyber risks evolve beyond human-scale response times.
The broader concern for regulators and executives is not only whether banks can identify vulnerabilities quickly, but whether they can repair them before malicious actors exploit the same weaknesses.
As AI capabilities continue to expand, cybersecurity may increasingly become a race between automated defense systems and equally sophisticated digital threats.
