OTSecurity

Manufacturing AI adoption doubles as data governance gaps, shadow AI and emerging cybersecurity threats rise, Netskope finds


New data from Netskope Threat Labs reported that AI (artificial intelligence) adoption across the manufacturing sector is expanding rapidly, creating new data exposure and cybersecurity risks as employees increasingly incorporate AI into business operations. Share of manufacturing users actively using AI applications rose from 40% to 67% over the past year, while organization-managed AI application use doubled from 35% to 70%. Microsoft Copilot ranked as the most widely adopted AI application, used by 90% of organizations, followed by Claude Platform at 89% and ChatGPT at 76%. However, 96% of employees use applications with embedded AI features, while 90% interact with systems that use customer or user data for model training, raising concerns about sensitive information exposure and compliance.

The November 2026 report, based on anonymized telemetry collected from Netskope One customers between June 1, 2025, and Aug. 31, 2026, also identified growing risks involving AI-connected systems and malicious AI-themed campaigns. Remote Model Context Protocol (MCP) activity increased sharply, with the number of non-human entities connecting to remote MCP servers rising 490% and MCP-related events growing 600%.

Regulated data accounted for 57% of AI-related data policy violations in manufacturing, followed by passwords and keys at 19%, intellectual property at 16% and source code at 9%. Netskope also observed malicious AI links reaching 200 interactions per 100,000 users in March 2026, highlighting the expanding threat of attackers exploiting AI applications and workflows to expose sensitive business information and distribute malicious content.

With AI now embedded across manufacturing, from managed applications to everyday software and agentic workflows, new security, data, and malware risks are following close behind.

Netskope identified that AI keeps spreading across manufacturing floors and back offices alike. Over the past year, active AI users rose from 40% to 67% of all users in the sector, and Microsoft Copilot now ranks as the most widely adopted AI application there, a pattern that sets manufacturing apart from other industries. Organizations are also tightening their grip on how the tools are used: managed AI application use climbed from 35% to 70%, while personal AI account use slipped from 75% to 55%. Yet AI activity inside these companies extends well beyond dedicated tools or direct usage. Some 96% of employees work with applications that carry embedded AI features, and 90% interact with systems that use customer or user data for training, a reality that compliance teams will need to factor into their planning.

Sensitive data, meanwhile, is flowing through these AI workflows in ways that raise real concerns. Regulated data accounts for 57% of AI-related data policy violations in manufacturing, followed by passwords and keys at 19%, intellectual property at 16%, and source code at 9%. Remote Model Context Protocol activity has also surged. The number of non-human entities connecting to remote MCP servers rose by 490%, and MCP events themselves grew by 600%. For organizations, this means the question is no longer just which AI applications and agents are running in their environments, but also which users and systems those agents interact with and what they are able to access.

Attackers, unsurprisingly, have noticed the trend and are working it to their advantage. AI-themed lures consistently target corporate users, with activity spiking in February and climbing further in June 2026. Malicious AI link interactions followed a similar trajectory, appearing in waves from July 2025 and rising again in early 2026.

Manufacturers have steadily reduced shadow AI risk by guiding users away from personal AI accounts and toward organization-managed tools. Over the past year, the use of personal AI applications fell from 75% to 55%, while adoption of organization-managed AI solutions rose from 35% to 70%. At the same time, the share of users switching between personal and enterprise accounts increased from 9% to 25%, suggesting that employees continue to move between managed and personal accounts as they explore new tools. Manufacturers can reduce this switching by speeding up the review and approval of new AI applications while keeping tighter instance-level controls.

Manufacturers are moving users to managed AI, improving data protection and compliance without limiting productivity. Personal AI use remains common, however, and the overlap between personal and enterprise accounts has grown. Shadow AI rarely disappears entirely, so organizations need long-term guardrails, clear policies, and a simple process for requesting new AI tools.

Netskope identified that many manufacturers are restricting specific AI applications because of security, privacy, and compliance concerns. While policies vary by organization, some applications are blocked more often than others, offering insight into where security teams see the greatest risks.

“Particular Audience is the most frequently blocked AI application in the manufacturing sector, with 45% of organizations restricting access. Emergent follows at 39%, while DeepSeek is blocked by 36% of organizations,” according to the Netskope report. “These restrictions reflect concerns about how AI applications handle sensitive information and connect to business systems; where prompts and uploaded files are stored; which internal resources an app can reach; and whether its processing happens outside established security and compliance controls.”

Across the manufacturing sector, 64% of employees use AI applications directly, while 96% use applications that include AI-powered features. In addition, 90% interact with AI systems that use customer or user data to train models, a compliance risk unless properly governed.

Most AI use happens inside applications employees already rely on, not in standalone tools. That makes it harder for manufacturers to track where sensitive information is shared and how it is used, both in direct interactions with AI tools and in AI features running in the background.

The Netskope report flagged how a malicious AI link is a harmful link returned by an AI application that a user clicks or an AI agent follows. “These links are similar to malicious links planted in search engine results, where users may click them believing they lead to legitimate websites. Attackers can promote them through SEO techniques, paid advertisements, or compromised legitimate infrastructure.”

“The approach is similar in AI environments, though malicious content reaches users through different channels. Techniques such as artificial intelligence engine optimization (AIEO) are emerging as attackers seek to influence the content AI models surface, while advertising is also beginning to appear within AI applications,” according to the Netskope report. “In the manufacturing sector, interactions with malicious AI links began to rise in July 2025, reaching 150 per 100,000 users in October. A second increase began in February 2026, with activity climbing further to 200 per 100,000 users in March. Security teams should inspect links returned by AI applications the same way they inspect links in search results and email, before users or agents follow them.”

Based on the trends uncovered in this report, Netskope Threat Labs strongly encourages organizations across the manufacturing sector to take a fresh look at their overall security stance.

The initial priority is to inspect all HTTP and HTTPS downloads, covering every stream of web and cloud traffic, so that malware cannot slip into the network unnoticed. Netskope customers can configure Netskope One NG-SWG with a threat protection policy that applies to downloads across all categories and all file types. Organizations should also block access to applications that serve no legitimate business purpose or pose a disproportionate risk. A sensible starting point is a policy that allows the reputable apps already in use while blocking everything else.

Data loss prevention (DLP) policies deserve a central place in this effort, since they can detect potentially sensitive information, including source code, regulated data, passwords and keys, intellectual property, and encrypted data, before it is sent to personal app instances, AI apps, or other unauthorized locations. Also, Remote Browser Isolation (RBI) technology can add a further layer of protection when employees visit websites in higher-risk categories, such as newly observed or newly registered domains.

The report also called on organizations to gain visibility and control over both managed and unmanaged AI applications and API interactions, applying real-time controls, consistent guardrails against sensitive data exposure and unsafe prompts, and governance policies that secure data flows between users, applications, and large language models. It further urged them to monitor AI adoption trends, user activity, and data loss incidents to better understand their risk exposure, and to consider red teaming private AI deployments so that vulnerabilities and misconfigurations are found before they can be exploited in production.



Source link