AI is transforming the world of work—and dramatically expanding the attack surface for cybercriminals. Proofpoint examines new risks such as prompt injections, AI-generated phishing attacks, and the abuse of trust relationships between humans and machines.
Artificial intelligence has firmly entered everyday working life with the availability of generative AI, leading to agent-based workplaces where humans collaborate, communicate, and make decisions together with AI agents. According to Proofpoint’s latest “AI and Human Risk Landscape Report,” 87% of companies already use AI assistants company-wide and 76% are introducing autonomous agents. However, only 48% have integrated security aspects into their AI strategy from the outset. This new form of collaboration offers enormous opportunities for productivity and efficiency, but also significantly expands the attack surface for threat actors. Moreover, AI-powered systems enable criminals to professionalize their activities and execute them at unprecedented speed and quality.Artificial intelligence has firmly entered everyday working life with the availability of generative AI, leading to agent-based workplaces where humans collaborate, communicate, and make decisions together with AI agents. According to Proofpoint’s latest “AI and Human Risk Landscape Report,” 87% of companies already use AI assistants company-wide and 76% are introducing autonomous agents. However, only 48% have integrated security aspects into their AI strategy from the outset. This new form of collaboration offers enormous opportunities for productivity and efficiency, but also significantly expands the attack surface for threat actors. Moreover, AI-powered systems enable criminals to professionalize their activities and execute them at unprecedented speed and quality.
New Attack Surfaces Through AI Tools
Where previously laborious manual work was required to create convincing phishing emails, generative AI tools can now produce deceptively real content in just minutes. These are linguistically flawless, perfectly matched in tone and structure to the target, and often include authentic brand elements. Fraudulent websites that steal credentials or even bypass multi-factor authentication can also be generated with AI tools. The technological barriers for such attacks are dropping dramatically. Even individuals without programming knowledge can create professional phishing sites thanks to AI. Proofpoint researchers observed hundreds of thousands of malicious URLs per month on the Lovable AI platform alone since February 2025; the number of apps with AI functionality rose from 11,290 in December 2024 to 258,033 in November 2025 — a 22-fold increase in one year.
A particularly insidious attack technique gaining traction in AI workplaces is the injection of hidden instructions into seemingly harmless content. These so-called prompt injections are concealed in emails or documents and take effect when an AI assistant processes the content uncritically. This allows dangerous commands to be executed unnoticed, security mechanisms to be bypassed, or confidential company data to be stolen. Attackers specifically exploit the trust employees place in the digital tools they use daily — email clients, collaboration platforms, or new AI copilots. 45.1% of security leaders cite AI manipulation through prompt injection as one of their biggest concerns.
Dr. Jakob Jung is Editor-in-Chief of Security Storage and Channel Germany. He has been working in IT journalism for more than 20 years. His career includes Computer Reseller News, Heise Resale, Informationweek, Techtarget (storage and data center) and ChannelBiz. He also freelances for numerous IT publications, including Computerwoche, Channelpartner, IT-Business, Storage-Insider and ZDnet. His main topics are channel, storage, security, data center, ERP and CRM.
Contact via Mail: jakob.jung@security-storage-und-channel-germany.de