Unit 42 frames ransomware and AI-driven attacks as one holiday retail risk

Unit 42 identified ransomware and AI-driven attacks as converged risks to holiday retail operations in 2026, requiring unified defenses against both traditional and emerging threats.

The always-on service uses OpenAI and Anthropic models for vulnerability detection and targets enterprises concerned about AI-driven cyber threats. It aims at security budgets seeking 24/7 offensive testing instead of periodic red teaming.#paragraph|{

The National Retail Federation predicted 2025 peak holiday sales would exceed $1 trillion USD for the first time, underscoring how profitable the season is for retailers and for cybercriminals. Salesforce reported that 2025 peak online sales equaled $1.3 trillion globally, including $294 billion in the United States and $262 billion influenced by AI and agents. Retailers face heightened risks from financially motivated attacks during peak seasons, including supply chain disruptions and operational downtime that threaten both enterprise and customer-facing operations. Unit 42 drew its peak-season attack priorities from years of observed activity against Fortune 100 retailers during peak seasons.

Ransomware operators targeted 565 retail organizations by Sept. 30, 2026, with median victim revenue of $22 million, underscoring the financial impact of ransomware on retailers during peak seasons. The top three ransomware cartels targeting retailers in 2026 are TGR-CRI-1173 (The Gentlemen), Spikey Scorpius (Qilin), and Howling Scorpius (Akira), all actively exploiting retail vulnerabilities during peak shopping seasons.

In December 2023, Ambitious Scorpius (ALPHV) stole millions of customer records from a retailer and caused delayed order fulfillment, canceled e-commerce orders, and slowed inventory replenishment. The incident demonstrates the operational risks of ransomware during critical retail periods, showing how a single attack can disrupt order processing and inventory restocking at the busiest time of year.

Unit 42 cites prior end-of-season disclosures that complicated retail security responses. The SolarWinds software supply chain compromise surfaced in December 2020 during a period when code freezes likely limited changes to critical technology systems. Attackers exploited the Log4j zero-day vulnerability in December 2021 during a period when code freezes likely limited changes to critical technology systems.

Extortion attacks without encryption dropped below 90% in 2025 for the first time in years, per Unit 42’s 2026 Global Incident Response Report, a shift that reflects evolving tactics in data theft and extortion campaigns. Bling Libra (ShinyHunters) and Hazy Scorpius (CL0P) increasingly use data theft and extortion instead of ransomware to speed up attacks, prioritizing speed and financial gain over encryption-based disruption. In December 2021, Bling Libra stole hundreds of gigabytes of customer data, including e-mail messages, hashed passwords, payment invoices, and employee files, highlighting the severe damage to customer trust from data theft.

Distributed denial-of-service attacks, which flood a target with traffic to knock it offline, pose a major risk to e-commerce platforms during peak seasons. In December 2025, the pro-Russian hacktivist group Noname057(16) used the DDoSia botnet to send coordinated TCP SYN and HTTP GET floods at web-facing portals and API endpoints. The attack took central systems offline, stopped package tracking tools, interrupted payment services for e-commerce merchants, and slowed parcel handling across retail distribution networks.

Attackers used the CosmicSting vulnerability, tracked as CVE-2024-34102 in the Common Vulnerabilities and Exposures catalog, in October 2024 to run remote code on unpatched servers, place persistent backdoors, and add payment skimming scripts to checkout flows. The digital skimming incident hit roughly 5% of online stores on prominent e-commerce platforms worldwide before Black Friday and Cyber Monday. Unit 42 cites digital skimming as an ongoing payment threat ahead of Black Friday and Cyber Monday.

In December 2022, attackers used automated scripts targeting mobile app and web login APIs with large databases of stolen credentials, compromising thousands of loyalty accounts. Credential stuffing, the use of stolen usernames and passwords to gain account access, let actors drain stored gift card balances, steal mobile payment funds, collect rewards points, and advertise compromised account access on dark web forums, showing how attackers monetize stolen credentials.

Retail customers typically increase gift card purchases during holiday periods, and cybercriminals ramp up targeting of this digital currency during peak season, tying seasonal consumer behavior to elevated fraud risk. Unit 42 reported the credit card fraud-based Jingle Thief campaign in October 2025 and continues to track it as CL-CRI-1032 (Storm-0539), an ongoing gift card fraud campaign tied to retailer employees. Actors behind Jingle Thief send phishing e-mail messages—fraudulent messages designed to steal credentials—and smishing text messages to trick employees at targeted retailers, use cloud-based infrastructure, and issue unauthorized gift cards at scale.

AI-powered attacks, including autonomous intrusions, can condense attack timelines to under 10 hours, as demonstrated in May 2026, marking a shift from manual attacks that required weeks to complete. Bling Libra used Anthropic’s Claude model to speed up reconnaissance and exfiltration in AI-driven attacks, showing how AI tools enable faster and more efficient attack execution. Unit 42 reported an incident in which a person used frontier AI models—advanced general-purpose AI systems—to independently handle parts of an intrusion in under 10 hours, a task that requires human operators at least two weeks.

Unit 42’s May 2026 prediction said threat actors would weaponize frontier AI models in 3–5 months, framing autonomous AI intrusion as a realized near-term threat for peak season. Gambit Security reported an attacker used open source harnesses to take more than 600,000 payment card records from retailers with JavaScript-based web skimmers. NorForge published findings that a threat actor used the Cairn autonomous penetration testing agent to infiltrate retail e-commerce sites, point-of-sale infrastructure, and web applications globally, demonstrating autonomous agent activity across retail assets worldwide.

To limit ransomware damage, retailers should deploy automated containment, network segmentation, and zero-trust access, measures critical to limiting lateral movement and operational disruption. They should also use endpoint protection to quickly isolate compromised assets with real-time behavioral analysis, and inline sandboxing to examine suspicious binaries and block zero-day exploits before they run, specific technical controls to limit ransomware impact on uptime.

Retailers must deploy data loss prevention controls and zero-trust access to prevent illicit data exfiltration, measures essential for protecting sensitive customer and transactional data. Organizations should combine data loss prevention, which blocks sensitive data from leaving the network, with continuous attack surface management to identify and remediate publicly accessible systems. Extortion incidents without encryption severely damage consumer trust in retailers safeguarding personal data, linking extortion volume to reputational harm during high-trust shopping periods.

Retailers must deploy dynamic rate-limiting, automated mitigation, and distributed edge networks to counter distributed denial-of-service threats, defenses that help absorb traffic spikes and protect critical systems. They must also deploy web application and API protection, which filters traffic to web applications and programming interfaces, and continuously scan for vulnerabilities to mitigate skimming risks, since proactive vulnerability management is critical to preventing skimming attacks.

Organizations should use cloud security posture management alongside application security posture management to continuously identify server vulnerabilities and correct configuration drift before holiday code freezes, supporting pre-freeze remediation to reduce skimming risk. Retailers must deploy web application and API protection with bot mitigation and advanced security to block credential stuffing and command-and-control traffic, the communications attackers use to control compromised systems, measures that help distinguish legitimate access from automated attacks.

Retailers should invalidate compromised accounts when defending against autonomous AI threats that can compress intrusion timelines to under 10 hours, a response action for AI-accelerated intrusions. They must unify endpoint, network, cloud, and identity data into automated security operations to defend against AI-driven threats, an approach that reduces mean time to detect and respond from hours to minutes.

Unit 42 recommends network segmentation, resistance to multifactor authentication fatigue—when repeated push prompts wear users into approving a login—point-of-sale and supply-chain visibility, and a customer-communication playbook as key priorities for retailers. These measures align with the converged threat of ransomware and AI-driven attacks during peak seasons.

Unit 42 points to its platform products for operational resilience during peak season. Retail security teams may use Cortex XSIAM and XDR to contain threats in real time together with Zero Trust network defense, and Cortex Cloud protects e-commerce platforms and defends critical transaction paths against high-impact threat vectors. Unit 42 Threat Intel Services and its Deep and Dark Web Service provide tailored adversary intelligence, direct analyst consultations, and specialized exposure tracking to organizations.

Frontier AI Defense includes three services: Frontier AI Exposure Analysis, Autonomous Security Blueprint benchmarking engagement, and Agentic Defense Transformation program. Unit 42 Frontier AI Exposure Analysis uses OpenAI’s GPT-5.6-Cyber model to identify and validate vulnerabilities in customer networks, a model that outperforms general-purpose AI in cybersecurity tasks. Unit 42’s multimodel harness assigns tasks to the best-suited AI model, broadening coverage of attack paths, with human review validating findings before remediation planning.#paragraph|{

Unit 42 Frontier AI Exposure Analysis uses OpenAI Group PBC frontier cyber models inside customer environments to find vulnerabilities, misconfigurations, leaked credentials, and unmanaged attack surfaces. Unit 42’s services integrate with existing IT, development, and security workflows to prioritize remediation of critical vulnerabilities before they are exploited, ensuring weaknesses are addressed ahead of attackers. Unit 42 said 36% of the exposures it identified have no known Common Vulnerabilities and Exposures, underscoring the novelty of exposures found by the service.#paragraph|{

Unit 42 consultants direct AI models and validate results using Palo Alto Networks telemetry and threat intelligence, keeping human review in the loop. Unit 42 runs adversary simulation against exposure findings to establish whether an exposure is exploitable and how far an attacker could travel inside customer networks. The remediation plan prioritizes fixes that break the most attack paths.#paragraph|{

Palo Alto Networks partners with Anthropic for early access to Claude Mythos Preview, a model specialized in finding software flaws, as part of a collaboration that aims to strengthen defenses against AI-driven exploits. Anthropic included Palo Alto Networks among roughly 50 organizations signed up in April for Project Glasswing. Claude Mythos Preview is not generally released because it is effective at discovering and linking software vulnerabilities.#paragraph|{

OpenAI expanded its Daybreak program on Aug. 10 with two access tiers, a partnership structure described in coverage of Palo Alto Networks’ participation. Daybreak Red uses the GPT-5.6-Cyber model for authorized vulnerability research and exploit validation. Palo Alto Networks was named among the partners in OpenAI’s Daybreak program.#paragraph|{

According to Sam Rubin, briefings have reached more than 1,000 security teams, and hundreds of customers were introduced to the service, adoption metrics cited by the Unit 42 executive. “We intend to spend it building on the side of the defenders,” said Sam Rubin, senior vice president of Unit 42 consulting and threat intelligence.#paragraph|{

With a market cap of US$321.7b, Palo Alto Networks relies on third-party frontier models instead of building everything internally, reflecting its global footprint in cybersecurity. The US-based cybersecurity provider counts enterprises and governments in the Americas, Europe, the Middle East, Africa, the Asia Pacific, and Japan among its customers.#paragraph|{

Yahoo Finance identified a signal for tracking the service’s revenue impact: whether Continuous Frontier AI Defense appears in Next Generation Security ARR and large platform deal commentary in fiscal 2027 and 2028 updates, including contracts over US$10m.#paragraph|{