Over half of manufacturers surveyed in a recent report experienced or suspected an AI agent security incident in the past year, following an ongoing trend of AI cybersecurity incidents and ransomware attacks on manufacturing.
Gravitee’s State of AI Agent Security Report, which surveyed 900 senior technology leaders across several sectors, found that only 22% of respondents gave every AI agent its own identity, making it harder to identify which agent took an action when shared credentials are used. Denver, Colorado-based Gravitee is vendor of an AI agent management platform.
See also: Stopping algorithmic civil war on the plant floor
Additionally, as AI agent deployment expands, the Gravitee report found across sectors that confidence in organizations’ ability to manage incidents has declined from six months ago, while the controls under their agents look the same. For example, respondents who said they were “very prepared” to respond to an agent incident fell from 47% to 29%.
In manufacturing, threat activity increased by almost 250% as of April 2026, which has largely been accelerated by AI both within and outside of manufacturing facilities. Manufacturers also remain a top target for ransomware attacks.
See also: Report: ‘Authentication friction’ hits manufacturing hardest, cybersecurity experts advise ditching passwords altogether
Manufacturing had the lowest average share of governed agents, at 45%, of all sectors that Gravitee surveyed. This includes financial services; healthcare; telecommunications; travel and transport; and the government and public sectors.
However, manufacturing did have the largest number of agents with their own identities, although that number was only 22%. Over 10% ran more than 150 AI agents, expanding the number of systems requiring security oversight, according to Gravitee.
See also: New Rockwell report reveals glaring disconnect between cybersecurity investments and risk protection
While 97% of respondents across sectors said they run agents that call other agents and 40% connect to other organizations’ agents, only 27% were able to define accountability formally across the full chain of agents. And 16% said it is unclear.
Broadly, the Gravitee report revealed significant gaps in organizations’ ability to control, monitor, and intervene in agent activity. Only 51% of agents had policy enforced on their actions; one in eight respondents could fully recover what an agent did 30 days ago; and only 6% could revoke a rogue agent’s access within one hour.
‘Consequences beyond data loss’
Despite gaps in agent governance and accountability, manufacturers are continuing to implement AI agents at increasing rates.
According to another new report, Netskope’s Threat Labs Report: Manufacturing 2026, active AI users rose from 40% to 67% of total users in manufacturing over the past year.
Netskope highlighted that for manufacturers, cyber incidents easily morph into physical and commercial threats. Plant operations and corporate IT are interconnected, so an intrusion can halt a production line and expose decades of intellectual property, resulting in consequences beyond data loss.
See also: U.S. agencies report cybercriminals used AI-generated code to crack Siemens PLCs
However, Netskope’s report also found that manufacturers have steadily reduced shadow AI risk by guiding users away from personal AI accounts and toward organization-managed tools. Over the past year, adoption of organization-managed AI solutions rose from 35% to 70%.
But at the same time, the share of users switching between personal and enterprise accounts increased from 9% to 25% as employees move between managed and personal accounts.
This echoes other recent findings that found manufacturing is hit the hardest by the consequences of “authentication friction,” partially due to the nature of shared workstations and supply chain interconnectivity.
Click Here For The Original Source.
——————————————————–
