Speaking at the NDTV World Summit 2026 in New Delhi, Butts, Microsoft’s Director of AI and Cybersecurity Policy, warned that artificial intelligence was increasing the capabilities of malicious actors, allowing them to execute sophisticated operations with greater speed, scale and automation.
Butts identified a China-based ransomware operation, espionage campaigns linked to state-sponsored groups and a commercially available cybercrime platform as examples of the changing threat environment. He said these developments demonstrated that cybersecurity risks were already materialising rather than remaining hypothetical concerns about future technology.
The first example involved Storm-3168, also known as Jadepuffer, which Microsoft tracks as a ransomware operation based in China. Butts said the group was employing AI agents to conduct attacks against organisations across international markets, illustrating how automation was transforming criminal operations.
Ransomware attacks typically involve criminals gaining access to computer systems, encrypting information or threatening its disclosure, and demanding payment. AI agents can assist attackers with multiple stages of an operation, potentially reducing the technical expertise and human intervention required.
The second threat concerned espionage activities involving groups linked to Russia and China. Butts said state-sponsored operators were incorporating AI agents into campaigns designed to obtain sensitive information from targeted organisations and networks.
He specifically identified Storm-0940, which Microsoft associates with China-linked activity focused primarily on data theft and espionage. The company had published fresh threat intelligence concerning the group on Thursday, a day before his remarks.
The distinction between criminal ransomware and state-sponsored espionage is significant because their objectives differ. While ransomware operators generally seek financial returns, espionage campaigns concentrate on obtaining confidential information, maintaining access to networks and gathering intelligence.
Butts did not disclose specific organisations in India compromised by the groups he identified, nor did he provide a country-by-country breakdown of their operations. The examples illustrated broader international threats relevant to India’s cybersecurity exposure.
His third example concerned EvilTokens, a cybercrime operation Microsoft began tracking in February. Butts said two individuals had used open-source and proprietary AI models to develop criminal infrastructure capable of operating with substantial autonomy.
According to Butts, the platform offered access to cybercrime tools for $1,500, alongside a monthly subscription of $500 for first-time customers. He said the operation had affected more than 10,000 organisations across 120 countries.
The figures highlighted how AI-assisted software development could enable small criminal operations to create and distribute capabilities that previously required considerably greater resources. However, Butts did not specify how many organisations affected by EvilTokens were located in India.
Microsoft’s assessment comes as its 2026 Digital Defense Report identifies AI as an increasingly important factor in the speed and accessibility of cyberattacks. The company says malicious actors are applying the technology to reconnaissance, phishing, malware development and activities following unauthorised network access.
The report also cautions against treating every AI-assisted intrusion as fully autonomous. Complex attacks still frequently involve substantial human direction, even as software agents take responsibility for individual tasks and accelerate operations.
Microsoft’s security research indicates that attackers continue exploiting compromised identities, phishing and weaknesses in access controls. AI can increase the effectiveness of those methods without necessarily changing the underlying vulnerabilities through which organisations are compromised.
Butts emphasised the scale of Microsoft’s monitoring operations to explain how the company identifies evolving threats. He said its systems processed approximately 165 trillion telemetry signals daily, screened around five billion emails and blocked roughly five million newly identified malicious files.
Those figures broadly correspond with Microsoft’s published cybersecurity measurements, which put daily email screening at 5.2 billion messages and malware blocking at approximately 4.7 million new files.
The company’s threat intelligence draws on activity observed across its security products and services. Its geographical rankings therefore reflect threats detected within Microsoft’s monitored environment rather than an exhaustive count of every cyberattack worldwide.
Microsoft’s 2026 report covers customer targeting patterns between July 2025 and June 2026. It also identifies government services, information technology, and research and academic institutions among the sectors experiencing substantial malicious activity.
The report records that 63 per cent of observed intrusions involved data theft, underscoring the importance of protecting sensitive information alongside preventing disruption to computer systems and digital services.
