Cybercrime is entering a dangerous phase as criminals increasingly exploit artificial intelligence to automate sophisticated attacks, steal sensitive information and execute extortion campaigns with minimal technical expertise.
Anthropic, developer of Claude, has documented how financially motivated hackers manipulated its AI systems to perform complex cybercriminal activities traditionally requiring experienced hacking teams.
In August 2025, Anthropic reported that an attacker used Claude Code against at least 17 organisations, including healthcare providers, a financial institution and a defence contractor.
The AI reportedly assisted with identifying vulnerabilities, developing intrusion tools, extracting confidential information and organising stolen financial, identity and defence-related records for extortion.
By September 2026, Anthropic identified further campaigns linked to ShinyHunters affiliates, demonstrating how AI-assisted cybercrime was becoming increasingly organised and operationally sophisticated.
One investigation uncovered an operation involving approximately 1.8 million Android application files analysed for exposed credentials. Another supply-chain compromise reportedly affected nearly 200 downstream customers within 34 hours.
This emerging phenomenon, described as "vibe hacking," fundamentally changes cybercrime economics. Criminals can provide broad objectives while AI agents undertake reconnaissance, generate code, analyse weaknesses and coordinate successive attack stages.
The implications for banking and financial services are particularly serious. Even when attackers compromise technology providers rather than banks directly, stolen payment credentials, identity documents and authentication tokens can enable financial fraud.
The biggest challenge is the growing asymmetry between attackers and defenders. AI-assisted intrusions can accelerate dramatically, while traditional security operations still depend heavily on manual investigation, fragmented tools and delayed responses.
Although Anthropic has reportedly disabled abusive accounts, strengthened safeguards and engaged law enforcement, restrictions cannot eliminate misuse across commercial and open-source AI ecosystems. Enterprises therefore need continuous identity verification, behavioural analytics, automated threat detection and stronger supply-chain security.
The strategic warning is unmistakable: cybersecurity must evolve from human-speed incident response to AI-powered, continuously adaptive defence. As autonomous agents lower the technical barriers to cybercrime, banks, insurers and enterprises must protect sensitive information wherever it resides. The next generation of cyberattacks may require fewer hackers, but deliver significantly greater scale, speed and financial consequences.
