Anthropic researchers used Claude Mythos Preview to find new weaknesses in two cryptographic methods, the company said Tuesday, including one that is being considered by the National Institute of Standards and Technology for both traditional and quantum computing.  

In a blog post detailing the work, the frontier AI company called it a “substantial” research advancement, but also emphasized that neither flaw affects software now in use.

“The attacks described in these two papers are the strongest attacks we have found to date,” the company wrote in the post. 

One of the weaknesses found was in HAWK, a digital signature scheme under review by the NIST as part of a search for encryption methods that could survive attacks from quantum computers. Working with a human researcher, the AI system found a mathematical shortcut, known as a nontrivial automorphism, in the lattice structure (a complex mathematical grid underpinning its security) HAWK relies on.

The discovered weakness cuts HAWK’s effective key strength in half, meaning key sizes would need to double to keep the same level of security. Anthropic said that change would erase much of what made HAWK an appealing candidate in the first place.

Ellen Boehm, senior vice president of strategy and AI innovation at Keyfactor, a digital identity and cryptography management provider, told CyberScoop that research like Anthropic’s proves that the NIST PQC evaluation process is working. 

She also said the research “elevates the importance for organizations to have visibility of where cryptography sits inside their enterprise, what business systems and processes it’s connected to, and the need for PQC readiness, if they haven’t already built a plan.” 

The other flaw was found in a weakened version of the Advanced Encryption Standard, or AES, the cipher NIST adopted in 2001 and the most widely used method for scrambling data in transit. Working largely on its own, Mythos invented a mathematical shortcut dubbed the “Möbius Bridge.” While real-world encryption scrambles data through 10 sequential layers, or “rounds,” researchers regularly study a simplified seven-round test version to measure security margins. In previous theoretical attacks, codebreakers had to check 256 separate values against a memory table, but Mythos created a shortcut that eliminated that lookup process entirely.

Combined with other optimizations, this discovery made the strongest known theoretical attack against seven-round AES 200 to 800 times faster. The attack is purely theoretical: It requires an impossible amount of target data — over 400 octillion messages — and cannot touch the full 10-round encryption protecting everyday software. Additionally, Anthropic pointed out that real-world systems remain completely safe.

Anthropic said it followed standard disclosure practices, notifying HAWK’s designers in June and coordinating public release with a NIST mailing list, and briefing government and industry partners beforehand. It also worked with researchers at ETH Zurich, Tel Aviv University and the University of Haifa to build a shared testing tool, called CryptanalysisBench, meant to let other researchers measure how AI systems perform against a range of ciphers.

The findings come as frontier AI models are being deployed by cybersecurity researchers in order to find vulnerabilities in all kinds of software. In June, intelligence agencies in the Five Eyes alliance warned that advanced AI models capable of wreaking havoc in the cyber domain are “months away.” However, a recent report found that despite the avalanche of bugs being unearthed, the threat level across the internet has not materially changed. 

Anthropic said it expects the same AI capabilities eventually to be applied to systems already in wide use, raising a separate question it said it has not yet resolved: how researchers, companies and governments should respond if a language model uncovers a flaw in a cryptographic system that protects critical infrastructure.

“As we develop increasingly powerful cryptanalytic results, it would be prudent to consider how researchers should react if a language model were to discover vulnerabilities in cryptosystems where attacks do have an immediate real-world impact,” the company wrote. “We hope that our work here will help launch these conversations.” 

Boehm said work like Anthropic’s further shows that enterprises should not rest on their laurels with any facet of their security apparatus. 

“AI is becoming a powerful tool for many things, including software quality assurance, code development, and in this case cryptographic analysis,” she told CyberScoop. “As AI tools become more widely and continuously used, it just elevates the need for enterprises to treat their trust infrastructure in an ongoing, operational manner versus thinking of it as a static environment that only changes every few years as new cryptographic algorithms are released.”

The post Here’s what Anthropic found when it turned Mythos loose on encryption algorithms appeared first on CyberScoop.

The head of a key federal government AI testing lab is leaving his post just months after taking over.

A Department of Commerce spokesperson confirmed to CyberScoop that Chris Fall is stepping down as director of the Center for AI Standards and Innovation, and his position is being backfilled.

“Following Chris’s departure, NIST Director Dr. Arvind Raman will continue to oversee CAISI and will serve as Acting CAISI Director,” the spokesperson said in a statement to CyberScoop.

Further details about the circumstances behind Fall’s departure were not provided. Axios, which first reported the departure, cited sources saying that Fall resigned.

The Center for AI Standards and Innovation has quietly become a key hub for the federal government to assess potential threats and harms that AI systems can pose to cybersecurity and national security. 

Early in the Trump administration, the center began informally working with frontier AI companies like OpenAI and Anthropic to test their models for threats, like their offensive hacking skills, assistance with building biological or nuclear weapons and other dangerous capabilities.

Fall was tapped to lead the center in April, and his departure just three months later comes as the White House has elevated the work of the center as one of the key means for determining which frontier AI models do — and do not — represent a step change in cyber or other capabilities compared to what’s available today.

Fall previously held other government posts, including as director of the Department of Energy’s Office of Science, assistant director for defense programs at the White House Office of Science and Technology Policy, and acting chief scientist at the Office of Naval Research.

The post Director of Commerce AI standards office out after three months appeared first on CyberScoop.

The AI-focused executive order President Donald Trump signed last month gave the Treasury Department, the National Security Agency, and the Cybersecurity and Infrastructure Security Agency (CISA) 30 days to establish a new “AI cybersecurity clearinghouse.” The deadline passed last week.

The clearinghouse is meant to coordinate the scanning, discovery, and validation of software vulnerabilities in critical infrastructure, and then prioritize how those vulnerabilities get patched and distributed.

It’s the right problem to solve. The question now is whether what is created will actually solve it.

The risk is that urgency produces something that looks like a clearinghouse, but functions like a committee: collecting information, convening meetings, and then stalling when it gets to the hard part.

Going beyond bug discovery is mission critical

It’s counterintuitive at a moment when AI-assisted vulnerability discovery is advancing rapidly, but the hard part is no longer just finding bugs. Those of us working at the intersection of AI and cybersecurity know where the real bottleneck is. HackerOne has seen it firsthand as a launch partner in Patch the Planet, OpenAI‘s initiative to use AI to find and fix vulnerabilities in critical open-source software at internet scale. The lesson underpinning that work, and informed by more than a decade of running vulnerability disclosure programs, is consistent: AI tools can surface vulnerabilities faster than anyone can act on them. What lags behind is everything that comes after discovery: deciding which findings are real, assessing severity in context, writing and testing a fix, and getting a patch accepted and deployed by the people responsible for maintaining the affected code.

Experienced human reviewers frequently disagree with AI-assigned severity ratings, because a model cannot see a project’s threat model or operational context. Software providers, especially the many volunteer open-source maintainers that so much of today’s digital infrastructure rely upon, face a relentless queue: verify the claim, assess the importance, write the patch, coordinate disclosure. AI has accelerated the incoming volume without yet equally accelerating our people and processes’ capacity to manage it. Better bug-finding tools mean you find more bugs. The improvements that really matter are the ones that help defenders push patches out and get them deployed faster.

That lesson should sit at the center of how the clearinghouse is designed.

If the clearinghouse focuses primarily on scanning coordination, which the executive order’s text emphasizes, it risks widening that gap rather than closing it. A body that finds more vulnerabilities but cannot move them to resolution is not a security win. At national scale, it is a backlog generator.

Laying a foundation for success

The administration can get this right, but it requires building the correct infrastructure now, not layering it on later.

The clearinghouse needs to do more than coordinate scanning. It needs to actually triage the results. Its core job should be filtering reports to identify which findings are truly credible, exploitable, and consequential for critical infrastructure. Using shared validation standards and risk-based prioritization, it can determine what warrants a national response. Otherwise, it’s just automating bigger backlogs.

Second, the clearinghouse also needs to tackle something more fundamental. Defenders don’t have the resources to respond to what gets reported. Vulnerabilities in critical infrastructure often live in open-source code maintained by small teams or individuals with no formal obligation to respond to disclosures and limited capacity to act quickly. The clearinghouse should work with the National Institute of Standards and Technology (NIST) to develop guidelines for open-source maintainers on structuring repositories and workflows to speed up patch review and deployment.

These guidelines should include how to use AI-assisted patching and clarify what downstream consumers of open-source code should do to help maintainers address vulnerabilities.  Federal policy should create incentives for downstream users to share responsibility for remediation through funding, engineering support, AI-assisted patch development, and procurement requirements that reward participation in coordinated vulnerability response.

Third, the clearinghouse should treat software bills of materials (SBOMs), the structured inventories of the components that make up a software product, as foundational infrastructure. SBOMs are what make it possible to trace where a vulnerable component lives across the supply chain. Without them, validated findings won’t be fixed fast enough at scale.

Finally, the clearinghouse should measure success based on what is fixed, not based on what is discovered.  Agencies need to publish data on validation rates, time-to-patch, adoption of fixes, and recurring classes of vulnerabilities. These metrics help AI systems, software vendors, and policymakers to continuously improve how vulnerabilities are addressed.

Most importantly: the agencies standing up this clearinghouse should resist the temptation to build its operational model from scratch. The private sector and the open-source security community have years of experience running exactly the kind of vulnerability intake, triage, and coordinated disclosure workflows the clearinghouse needs. The executive order wisely calls for voluntary collaboration with industry. That collaboration should be structural, not advisory, embedded in how the clearinghouse operates from the start, not bolted on after the architecture is already set.

The clearinghouse can work. But the challenge is no longer finding vulnerabilities. It is building a system that can turn discoveries into action. That is how its success should be measured.

The post Found fast, fixed slow: The gap the AI clearinghouse must close appeared first on CyberScoop.

A Department of Commerce inspector general report released Thursday found that the National Institute of Standards and Technology has mismanaged a critical cybersecurity vulnerability database through poor planning, inefficient operations, duplicate federal programs, and failure to communicate with users.

The National Vulnerability Database, maintained by NIST since 2005, collects information about computer security flaws and adds details like severity ratings and affected products. This information helps cybersecurity professionals across government and the private sector decide which security problems to fix first. In February 2024, the database’s enrichment contract lapsed, creating a backlog of unprocessed security flaws that has only grown worse.

The report identified the lack of strategic planning as a core problem. NIST leaders admitted they had no long-term plan for clearing the backlog, even as it grew from about 13,000 unprocessed security flaws in June 2024 to over 27,000 by the end of 2025.

NIST publicly promised in May 2024 that it would clear the backlog by September 2024, setting a goal of processing 6,200 security flaws per month, but the agency had never processed more than 5,000 per month in the past.

The report found major inefficiencies in how NIST enriches the information that is attached to the vulnerabilities. 

Analysts spend about 80% of their time on two tasks: calculating severity scores and identifying which products are affected. The inspector general’s office tested NIST’s severity scores and found they matched independent evaluators only 12% of the time. Also, nearly 80% of vulnerability submissions already include these scores from the companies that are responsible for the software. This means NIST is doing work that is often unnecessary and inconsistent. The inspector general proposed cutting back on severity score calculation work over the next two years, estimating that NIST would save $800,000 that it could redirect to other program areas.

Another efficiency problem highlighted is the program’s manual process for identifying affected products. Creating these standardized product identifiers takes a lot of time and keeps analysts from clearing the backlog. NIST is developing tools to make this faster, but it remains a major slowdown.

The report also found major duplication between two federal security programs. When the Cybersecurity and Infrastructure Security Agency launched its own Vulnrichment program in May 2024, there was no coordination between the agencies, leading to NIST analysts sometimes repeating work that CISA analysts had already completed. Additionally, the two agencies even hired the same contractor for portions of the same work. The inspector general found at least 21,000 cases of duplicated work between May 2024 and December 2025, wasting approximately $200,000 in the process.  

Communication failures have made the problems worse. In April 2024, over 50 cybersecurity professionals sent an open letter to Congress complaining that NIST was not being transparent about the database’s problems. Neither NIST nor the Department of Commerce answered the letter.

Vulnerability database programs managed by the federal government have been a point of contention for the cybersecurity community over the past two years. Earlier this year, NIST announced that it has narrowed its priorities for the NVD, focusing only on vulnerabilities in CISA’s KEV catalog, software used by the federal government, and critical software identified under Executive Order 14028.

A similar program that serves as a catalog of known security flaws, the Common Vulnerabilities and Exposures (CVE) list, has had similar issues over the past few years. That program, run by CISA, narrowly escaped a sudden demise when a last-minute, 11-month contract extension averted a shutdown in April 2025. Since then, several competing databases from European nonprofits and other private entities have been stood up in order to better coordinate how vulnerabilities are tracked, disclosed, and ultimately patched.

The inspector general recommended that NIST create a long-term plan for the database, set up a plan to clear the backlog with specific goals, cut back on unnecessary severity score work, make it easier for outside companies to help identify affected products, immediately start working with CISA to stop duplicating work, and develop a plan to communicate better with users.

NIST agreed with all six recommendations and said it is working on them. The agency must submit a plan showing how it will address these problems by late July.

You can read the full report here. 

The post Federal audit reveals NIST’s NVD is plagued by poor planning and duplication appeared first on CyberScoop.

Today’s enterprise executives are navigating a complex landscape of AI-driven challenges, but none is more urgent than the rapid escalation of AI-generated fraud.

Fraudsters are weaponizing generative AI to automate impersonation and mass-produce synthetic identities at a scale and pace that is rendering enterprises’ long-standing defenses obsolete. This is no longer a slow-moving game of cat and mouse; it is a high-velocity arms race.

To protect the integrity of their platforms, enterprise leaders — particularly in critical infrastructure sectors — must move beyond periodic risk assessments and begin leveraging a new generation of tools that enable defenses to iterate in days rather than months.

Generative AI as a fraud multiplier

While legitimate businesses use generative AI for efficiency, fraudsters exploit it to scale their attacks. We are witnessing a 100-fold increase in synthetic identities and a sevenfold rise in deepfake-driven impersonations over the past 24 months. Deloitte’s Center for Financial Services predicts AI-enabled fraud losses could reach $40 billion in the U.S. by 2027, up from $12.3 billion in 2023.

This is no longer just a back-office technical issue; it has become a top concern for leadership across banks, fintechs, and telcos. Three-quarters (72%) of business leaders anticipate AI-generated fraud, including deepfakes, will be a top operational challenge in 2026, according to an Experian report. Nearly half (46%) of businesses surveyed by Incode in 2025 reported an annual increase in deepfake and generative AI fraud.

Bad actors can now perpetrate fraud at scale by targeting multiple victims at the same time using the same or fewer resources. Consequently, the stakes have escalated rapidly. Enterprises must now find more effective ways to distinguish between reality and fiction before these attacks compromise trust, revenue, and operational continuity.

The new arms race

Fraud prevention has always been a constant game of leapfrog. Now, however, enterprises must adopt highly advanced defenses as they work to thwart fraudsters who have access to the same AI tools and no legal guardrails.

By some estimates, 80% of fraud is easily detectable, while the remaining 20% requires high-level expertise. That’s where most vendors’ performance fails. Sophisticated fraudsters are not only more capable of impersonating identities but are also increasingly networked, sharing intelligence on how to bypass specific company defenses.

Agility as the primary security metric

In this environment, the “7-Day Benchmark” is essential. A defense model must be able to identify a new attack vector, retrain its data sets, and deploy an updated mitigation model within 7 to 10 days. 

One reason so many organizations remain vulnerable to this new generation of attacks is that they rely on third-party vendors whose update cycles can take months to test and deploy. Modern defense requires an approach like Deepsight: a combination of machine learning, behavior checks, and device checks that identify camera injections and synthetic document fraud and verify that the user is a real person.

Defense checklist: 4 questions for every vendor

To narrow this “velocity gap,” executives need to take a closer look at how well equipped their providers are to address this new generation of threats.  Here are four pointed questions to explore:

1. “How accurate is your facial recognition capability? And what third-party certifications do you hold for mobile environments?” Executives should look for solutions that have been independently validated against the most rigorous international standards for biometric spoof testing—such as iBeta Level 3 compliance on both iOS and Android—that simulate well-resourced attackers using professional-grade, hyper-realistic masks.
   • While many providers struggle with consistency across various devices, a top-tier solution will achieve a 0% error rate. (In a 2024 National Institute for Standards and Technology (NIST) evaluation of 158 different developers, using galleries of mugshot, Visa, and Border images, Incode ranked #1 out of all full solution identity verification providers.) 
   • Also, assess the accuracy and performance of algorithms used in facial analysis across a range of use cases, including age estimation, ensuring the technology is unbiased and highly accurate across diverse user populations. (Once again, Incode scored top marks in NIST’s Face Analysis Technology Evaluation for achieving the lowest error and false-positive rates.)
2. “How do you measure and report your own error rates?” Demand a rigorous, audited approach that provides clear metrics on false positives and false negatives for every session.
3. “Do you own your technology or license it?” This determines the speed of iteration. Updates should happen internally in days, not over months-long development cycles dictated by a third party.
4. “How does your network share intelligence to flag repeat offenders?” Inquire whether the vendor can cross-share biometric, VPN, and network data across their entire client base to proactively block known fraudsters before they hit your system.

(For a more complete guide on selecting an identity verification vendor, we recommend getting a complimentary copy of the Gartner Magic Quadrant for Identity Verification.)

Secure your defenses against AI-enabled fraudsters

The era of treating identity verification as a static compliance checkbox is over. As the internet makes identity spoofing easier than ever before, the burden is on leadership to ensure their defenses can evolve at the speed of the adversary.

Audit your vendor ecosystem today: Demand proprietary technology that iterates in days, insist on top-tier independent certifications for mobile environments, and prioritize networks that share real-time intelligence. Organizations that treat trust as a core strategic capability will thrive; those that remain reactive will find themselves increasingly vulnerable in a world where reality is becoming ever more malleable.

Fernanda Sottil is Senior Director of Strategy at Incode Technologies, a leading identity verification company.

Learn more: Find out how Incode helps leading organizations eliminate fraud before it happens.

The post Weaponized AI: The new frontier of fraud and identity spoofing appeared first on CyberScoop.

The average cyberattack costs for a small- or medium-size business is more than $250,000. The salary for a chief information security officer (CISO) is about the same, pulling in between $250,000 and $400,000, according to the annual 2026 CISO Report from Sophos and Cybersecurity Ventures. Small- and medium-size businesses (SMBs) know they cannot afford the salary, so they roll the dice, hoping they will not be attacked. This is a dangerous gamble that these businesses, which make up the backbone of the American economy, should not have to take. A virtual (vCISO) or fractional CISO (fCISO) can provide a practical solution.

As the American economy goes digital, SMBs now rely on the same building blocks as big enterprises — cloud services, payment systems, remote access, customer data, and other third-party vendors.  But without senior cyber leadership, cybersecurity often becomes a patchwork of tools, checklists, insurance paperwork, and whatever guidance a vendor offers. That may get these companies through a questionnaire; it will not build real resilience. Nearly half, all reported cyber incidents, which is projected to cost the global economy $12.2 trillion annually by 2031, involve smaller firms.

The threat is growing in both size and sophistication. Adversaries are deploying AI to automate reconnaissance, develop malware, and run phishing campaigns at scale.  This reduces the cost and skill needed to target smaller firms at volume. Adversaries are also collecting encrypted data with the intent to decrypt it later when they have access to large enough quantum computers. SMBs in defense, healthcare, and financial supply chains often hold sensitive credentials that provide access into larger enterprise environments, but most are not prepared to adopt quantum-resistant encryption.

SMBs generally understand they face cyber risk. The real gap is leadership: someone who can turn technical vulnerabilities into business decisions, set priorities, brief executives, prepare for audits, and hold vendors accountable. For more SMBs, hiring a full-time CISO is financially unrealistic.

A Virtual CISO provides remote, on-demand cybersecurity leadership and advice, typically supporting several organizations at the same time. A fractional CISO is a dedicated, part-time executive who is more deeply integrated into one organization’s governance, security planning, and day-to-day operations. Both models give smaller organizations access to senior-level cybersecurity expertise in a flexible, more affordable way than hiring a full-time CISO.

Washington should make it easier for SMBs to hire fractional cybersecurity leaders, because the private market is not closing this gap on its own. The Cybersecurity and Infrastructure Security Agency (CISA) and the Small Business Administration (SBA) could help by publishing buyer guidance: vetted criteria for evaluating providers, example scopes of work and deliverables, and real-world case studies that show SMB owners what a high-quality vCISO or fCISO engagement should look like.

Clear guidance matters because many smaller firms cannot easily tell the difference between true cybersecurity leadership and a tool reseller, compliance-only consultant, or a generic managed services contract. Any vetted provider criteria should emphasize proven experience building and running security programs, independence from vendor incentives and product quotas, and the ability to tie security investment to real business risk, not just a list of certifications. Model scopes of work should also spell out the basics every engagement should deliver: an initial risk assessment, a prioritized remediation roadmap, and simple metrics that show whether security is improving over time. Without clear buyer criteria, federal efforts could end up funding low-quality services that add cost and paperwork without making companies safer.

The National Institute for Standards and Technology (NIST) should recognize these CISO models in its SMB-focused Cybersecurity Framework guidance. That would help smaller firms turn the framework’s Govern, Identify, Protect, Detect, Respond, and Recover functions into a clear, accountable leadership structure. This would make these roles less abstract: the point is not merely providing advice, but taking executive-level ownership of risk priorities, vendor oversight, incident readiness, and communication with the owner or board.

Congress and the Treasury Department should consider targeted tax incentives or credits for qualified cybersecurity leadership services, tied to measurable risk-reduction outcomes. Eligible activities could include completing a risk assessment, building a incident response plan, conducting vendor security reviews, running employee training, and producing a remediation roadmap. SMBs often defer cybersecurity because every dollar competes with payroll, inventory, and growth. A targeted incentive would make security leadership easier to justify as a business investment rather than an optional add-on.

Federal acquisition officials should require contractors that handle sensitive government data to show it has executive-level cybersecurity oversight, whether it is full-time, virtual, or fractional, and should extend that expectation down to relevant subcontractors and suppliers. This is necessary because SMBs serve as entry points into defense, healthcare, financial, and critical infrastructure supply chains.

Finally, CISA and the SBA should support vCISO- and fractional-CISO-led workforce training. Employees improve security when training comes with leadership, regular reinforcement, and clear accountability, not just annual awareness training. The aim is not to turn every SMB into a Fortune 500 security shop. It should be to give smaller firms access to the leadership they need before the next incident forces the issue.

Georgianna Shea, who is a Doctor of Computer Science, is chief technologist at the Foundation for Defense of Democracies’ Center on Cyber and Technology Innovation and its Transformative Cyber Innovation Lab, where Cason Smith served as a summer 2025 intern. Cason is studying integrated information technology at the University of South Carolina.

The post The missing cybersecurity leader in small business appeared first on CyberScoop.