Incident Overview
On December 31, 2025, Koi Security’s AI-assisted Wings platform generated a threat intelligence report falsely associating MeetingTV’s legitimate domains with malicious infrastructure linked to China. The report, published without human validation, led downstream cybersecurity vendors and network providers to block MeetingTV’s services, disrupting its operations. MeetingTV, an online videoconferencing and webinar provider, only discovered the issue after its domains began being flagged and blocked, with no prior notification or opportunity to dispute the findings.
The dispute escalated into litigation in March 2026, with MeetingTV filing a lawsuit against Palo Alto Networks (Koi Security’s parent company) and Koi Security itself. The lawsuit alleged that the AI-generated report contained ‘hallucinated findings,’ falsely implicating MeetingTV in Chinese state-sponsored espionage. Public awareness of the incident grew in late June 2026, following media coverage and the release of legal documents detailing the allegations. The incident was formally recorded in the AI Incident Database (AIID) on July 14, 2026, under Incident ID 1587.
The false classification of MeetingTV’s domains as malicious resulted in three primary harms: service disruption and denial of service, reputational damage, and the spread of misinformation. Customers relying on MeetingTV’s platforms experienced interrupted access to its services, while the company faced reputational risks due to the unfounded association with state-backed cyber threats. Additionally, the incident highlighted broader concerns about the reliability of AI-generated threat intelligence when deployed without rigorous human oversight.
Koi Security’s Wings platform, an AI-assisted tool designed to analyze and classify cyber threats, played a central role in the incident. While AI tools are increasingly used to accelerate threat detection, this case underscores the risks of over-automation, particularly when human review processes are bypassed or insufficiently robust.
The incident aligns with a growing trend of AI-generated threat intelligence producing false positives that harm legitimate organizations. MeetingTV’s legal action against Koi Security and Palo Alto Networks marks one of the first instances where a company has sought redress for damages caused by an AI-driven security report, setting a precedent for future disputes in the cybersecurity industry.
Possible Root Cause
Koi Security’s report falsely associated MeetingTV’s domains with a Chinese state-sponsored hacking group, likely due to superficial similarities in network patterns, IP ranges, or domain registrations.
A critical factor contributing to the error was the absence of adequate human review before the report’s publication. Koi Security’s process appears to have prioritized speed and automation over verification, allowing the AI-generated findings to be disseminated without thorough validation. This approach is particularly risky in threat intelligence, where false positives can have immediate and far-reaching consequences for targeted organizations. The lack of a dispute resolution mechanism or pre-publication verification step further exacerbated the issue, as MeetingTV had no way to challenge the report before its effects materialized.
Without contextual understanding or domain-specific expertise, the AI lacked the nuance required to differentiate between legitimate operations and genuine threats.
Additionally, the downstream adoption of Koi Security’s report by other cybersecurity firms and network providers amplified the harm. Once MeetingTV’s domains were flagged as malicious, automated systems across the industry acted on the classification without independent verification, leading to widespread service disruptions. This highlights the need for greater transparency and accountability in threat intelligence sharing, particularly when AI is involved.
Takeaways
While AI tools can enhance threat detection by processing vast amounts of data at scale, their outputs must be subjected to rigorous human oversight to prevent false positives that harm legitimate entities. Organizations deploying AI-assisted security systems should implement multi-layered validation processes, including peer review, domain expertise, and dispute mechanisms, to mitigate the impact of automated errors.
For companies like MeetingTV, the case underscores the importance of proactive monitoring of threat intelligence feeds. Organizations should regularly audit their domains and infrastructure against public blocklists and security reports to identify and contest misclassifications before they result in service disruptions. Additionally, legal recourse may be necessary in cases where AI-generated reports cause demonstrable harm, as MeetingTV’s lawsuit illustrates.
The broader cybersecurity community should take note of the incident’s implications for trust in AI-driven threat intelligence. False positives not only disrupt legitimate services but also erode confidence in security tools and the organizations that deploy them. Transparency in AI decision-making processes—such as disclosing the confidence levels of automated findings—could also help users assess the reliability of threat reports.
Regulators and industry bodies may need to address the growing issue of AI-generated misinformation in cybersecurity through standardized guidelines or certification requirements. Similarly, establishing a framework for disputing and correcting false classifications could reduce the collateral damage caused by automated errors. The OECD AI Incidents and Hazards Monitor (AIM) and similar initiatives play a role in tracking such incidents, but proactive measures from vendors and policymakers are essential to prevent future occurrences.
Initiatives that encourage peer review, independent validation, and open dialogue about AI limitations could help mitigate risks. As AI continues to reshape cybersecurity, balancing innovation with responsibility will be key to ensuring that automated systems enhance security without inadvertently causing harm to legitimate actors.
Sources
Incident summary from the AI Incident Database, used under CC BY-SA 4.0.