Ivanti is experimenting with large language models to accelerate vulnerability remediation, according to the company’s chief security officer, Daniel Spicer. The effort reflects a broader push across the security industry to apply frontier AI models to the labor-intensive work of triaging and fixing software flaws.

Spicer said early results have been encouraging, particularly in the initial stages of the remediation workflow, where LLMs can help parse vulnerability data, correlate findings, and suggest fixes faster than manual processes allow. That speed could matter for organizations racing to patch flaws before attackers weaponize them.

Open Questions Remain

Despite the promising early performance, Ivanti’s leadership cautioned that significant questions remain before LLM-assisted remediation can be trusted at scale. Two issues stand out:

  • Cost: Running frontier models at the scale needed for continuous vulnerability management across large environments carries substantial compute expense, raising questions about return on investment compared to traditional tooling.
  • Human-in-the-loop viability: Security teams still need to determine how much human review is required to safely validate AI-generated remediation steps, especially for changes that touch production systems or critical infrastructure.

These concerns echo a pattern seen elsewhere in the industry as vendors explore generative AI for security operations. While LLMs can accelerate the early, data-heavy phases of vulnerability response, such as identifying affected assets or drafting patch recommendations, most organizations remain cautious about removing human oversight from decisions that could break production systems or introduce new risk.

Ivanti has not detailed a timeline for broader deployment of LLM-assisted remediation tools, and the company’s comments suggest the technology is still in an evaluation phase rather than production use. For security teams watching the space, the takeaway is that generative AI may soon play a larger role in vulnerability management, but cost controls and validation safeguards will likely determine how quickly that shift happens.