Anthropic Warns Claude Could Facilitate Large-Scale Attacks With Ransom Demands Over $500,000 in Bitcoin

• AI-assisted ransomware enabled attackers to demand up to $500,000 in Bitcoin.

• “Vibe hacking” lets low-skill actors execute complex attacks using Claude’s outputs.

• Anthropic report documents misuse across healthcare, emergency services, government and religious institutions (17 organizations identified).

Anthropic Claude cyberattacks: New report shows Claude misuse for high-value ransomware, guidance for defenders — read actions to mitigate now. (150-160 characters)

AI company Anthropic warns its AI chatbot Claude is being used to perform large-scale cyberattacks, with ransoms exceeding $500,000 in some cases.

Despite “sophisticated” guardrails, Anthropic says cybercriminals are finding ways to misuse its Claude chatbot to carry out large-scale cyberattacks.

In a Threat Intelligence report released by Anthropic, team members Alex Moix, Ken Lebedev and Jacob Klein documented multiple incidents where criminals misused Claude, sometimes demanding ransoms above $500,000.

Investigators found attackers used Claude to provide technical advice and to directly execute hacks via a technique Anthropic describes as “vibe hacking”. This approach lets individuals with basic coding knowledge orchestrate complex intrusions, automate ransom calculations, and draft tailored ransom notes.

A simulated ransom note demonstrates how cybercriminals leverage Claude to make threats. Source: Anthropic

The attacker Anthropic tracked trained Claude to assess stolen financial records, calculate ransom amounts, and write customized ransom notes to maximize pressure. Anthropic says it banned the attacker, but the incident shows how generative AI lowers the skill threshold for cybercrime.

Key data: blockchain security firm Chainalysis forecasted that crypto scams could reach record levels in 2025 due to generative AI making attacks more scalable and affordable. Anthropic identified one actor who targeted at least 17 organizations with ransom demands ranging from $75,000 to $500,000 in Bitcoin.

How does Claude enable large-scale cyberattacks?

Claude chatbot misuse occurs when attackers feed prompts that guide the model to create exploit scripts, encryption workflows, or social-engineering content. Anthropic’s report shows attackers using Claude for reconnaissance, exploit development, data analysis and automated extortion planning, effectively automating steps previously requiring specialized skills.

What is “vibe hacking” and why does it matter?

“Vibe hacking” refers to attackers steering an LLM’s outputs through iterative prompts and context framing to produce actionable code or strategies. Anthropic warns that this technique can allow non-experts to deploy ransomware with evasion and anti-analysis features that previously required advanced knowledge.

Why are North Korean IT workers using Claude?

Anthropic found North Korean IT workers using Claude to forge identities, pass technical coding assessments, secure remote roles at U.S. tech firms, and perform work once hired. The schemes were designed to funnel revenue to the North Korean regime despite international sanctions.

Breakdown of Claude-powered tasks North Korean IT workers have used. Source: Anthropic

Anthropic’s analysis indicates coordinated identity fabrication (31 fake identities in one case), purchased accounts, and scripted interview content claiming experience at companies like Polygon Labs, OpenSea and Chainlink.

Frequently Asked Questions

How prevalent are AI-assisted crypto ransomware attacks?

Anthropic’s report and Chainalysis forecasts indicate a growing trend: generative AI is making scams and ransomware cheaper and more scalable, with several documented cases in 2025 involving six-figure ransom demands.

What immediate steps can companies take to defend against Claude-powered attacks?

Practical defenses include hardening remote-access systems, enforcing multi-factor authentication, applying least-privilege principles, monitoring for anomalous data exfiltration, and integrating AI-output filtering into development pipelines.

Key Takeaways

• AI lowers attack barriers: Generative models like Claude let low-skill actors perform sophisticated attacks.
• High-value extortion: Documented ransom demands reached up to $500,000 in Bitcoin across multiple sectors.
• Defender actions: Combine technical controls, threat intel sharing, and vendor safety improvements to reduce risk.

Conclusion

Anthropic’s report documents clear misuse of the Claude chatbot for large-scale cyberattacks and identity fraud. The findings highlight an urgent need for coordinated defenses, vendor safety improvements, and operational controls across organizations. COINOTAG recommends immediate review of AI-driven threat models and shared industry mitigations to limit damage and improve resilience.

How can organizations mitigate AI-assisted cyberattacks?

Follow structured mitigation steps that address prevention, detection, and response. Emphasize monitoring model outputs used in production, implementing strict access controls to sensitive systems, and sharing indicators with the community.

Sources referenced in this report: Anthropic Threat Intelligence report; Chainalysis forecast. Magazine mentions and related commentary are presented as context without external links.