N-able, Inc. (NYSE: NABL) released a research report finding that 75.5% of organizations do not fully trust AI-powered security remediation without human review, even as AI tool adoption accelerates across the industry.

The report, titled "The AI Fog Report: When the Signal Becomes the Noise," is based on a survey of 595 cybersecurity decision-makers and practitioners in the United States and United Kingdom. The findings identify trust gaps, alert overload, and governance shortfalls as key challenges facing security teams.

According to the report, 86.1% of respondents said security alert volumes increased over the past 12 months, with 24.2% reporting a rise of more than 50%. The leading causes cited were adding more detection tools (40.2%) and AI enabling faster attacker activity (26.2%). Additionally, 54.8% of respondents said alert noise interferes with incident response daily or several times per week.

The survey also found that 84.5% of organizations encountered a confirmed or suspected AI-assisted attack in the past year. The leading sources of low-quality alerts were duplicate alerts from overlapping tools (50.4%) and AI-generated alerts lacking sufficient context (48.9%).

On the adoption side, 81.0% of respondents said they meaningfully increased AI-powered security tool adoption in the past year, with another 16.5% planning to do so in the next 12 months. Despite this growth, only 24.5% fully trust autonomous AI remediation. Before extending that trust, 62.7% require full explainability and 59.8% require a built-in human override capability.

The report also found that 87.1% of respondents said their organization made at least minor security policy exceptions to enable AI tools, and only 51.9% maintain a comprehensive inventory of AI tools operating across their environments.

"Security teams are entering the AI era from an already difficult operating position," said Brendan Griffin, Director of Threat Research at N-able. "Simply adding more AI doesn't automatically produce better security outcomes."

Only 41.0% of respondents said they are very confident their existing security stack can keep pace with AI without a significant overhaul.