AI Now Carries Out Cyberattacks With Little Human Input: Check Point Report
Source: The Straits Times
For the first time, AI tools have automated significant parts of cyber intrusions over the past 12 months, performing in minutes what used to take skilled attackers hours or days, according to a new 56-page report from Check Point Research.
Artificial intelligence has crossed a critical threshold in cybersecurity: AI systems can now autonomously execute significant portions of cyberattacks with minimal human intervention, a landmark report from Check Point Research has found. The 56-page study, released on July 14, documents cases where AI handled nearly every phase of an intrusion — from social engineering and malware development to live network penetration and vulnerability research.
Check Point Research, the intelligence arm of global cybersecurity firm Check Point Software Technologies, identified several real-world instances of AI-driven attacks. In one case, a single attacker used Anthropic’s Claude Code alongside OpenAI’s GPT-4.1 to breach nine Mexican government agencies, exfiltrating 400 million records across tax, civil registry, and healthcare databases. The campaign ran from late December 2025 to mid-February 2026, with AI automating the majority of the intrusion workflow — a task that would have previously required a coordinated team of skilled operators.
In a separate campaign, a Chinese-linked threat group used Claude Code to automate up to 90 percent of an espionage operation targeting 30 organisations globally across technology, finance, chemicals, and government sectors. Anthropic disclosed the breach in November 2025, marking the first known case of a cyberespionage campaign driven largely by an AI system. The AI scanned networks, identified vulnerabilities, stole credentials, and moved across compromised infrastructure, while human operators mainly set objectives and reviewed the output at key decision points.
The report also highlights a sharp rise in workplace AI-related risk. Between October 2025 and May 2026, organisations used an average of 10 different AI applications per month, while employee AI tool usage surged 25 percent. Critically, the proportion of high-risk prompts — those containing confidential corporate or personal data shared with external AI tools — doubled from 2 to 4 percent over the same period. Asia-Pacific recorded the lowest rate at 2.88 percent, but the upward trend signals growing exposure.
Why it matters for Singapore: As one of Asia’s most digitised economies and a regional cybersecurity hub, Singapore faces heightened exposure to AI-enabled threats. Check Point has a significant presence in Singapore, and the report’s findings serve as a critical reminder for local enterprises and government agencies to strengthen AI governance, adopt continuous monitoring of employee AI tool usage, and put safeguards around AI models before deployment. The era of AI-powered, low-cost, high-speed cyberattacks is here — and Singapore’s defenders must adapt just as fast as the attackers.