#USA #Austin #AI Security #agentic AI #HiddenLayer

HiddenLayer's 2026 AI Threat Landscape Report

On March 18, 2026, HiddenLayer revealed its AI Threat Landscape Report for 2026, detailing the current landscape of AI security challenges, particularly focusing on the rise of agentic AI—systems capable of autonomous actions. As companies deepen their integration of AI into critical operations, the report highlights a growing vulnerability within enterprise security frameworks. Surveys conducted among 250 IT and security leaders paint a concerning picture of the challenges organizations face.

Agentic AI and Emerging Threats

The report indicates that agentic AI is linked to 1 in 8 reported AI breaches. This alarming statistic underscores that as organizations transition from AI as experimental tools to fully autonomous agents, their security controls are struggling to keep up. Chris Sestito, CEO and Co-founder of HiddenLayer, emphasizes the rapid evolution of agentic AI, which has advanced more in the past year than many security programs have in five. As these agents gain capabilities such as browsing the web, executing code, and carrying out complex workflows, they introduce novel vulnerabilities for exploitation.

Key Findings from the Report

Some notable insights from the report include:

1. Expanding Exposure to AI Supply Chain Risks

The threat landscape is further complicated by malware typically hidden in public model repositories, named as the primary source of AI-related breaches by 35% of respondents. This dilemma illustrates the trade-off many organizations face between operational speed and security.

2. Gaps in Visibility and Transparency

A concerning 31% of companies cannot confirm whether they have faced AI security breaches in the last year. Despite the majority advocating for mandatory breach reporting, more than half admitted to withholding such disclosures out of fear. This behavior reveals a stark contradiction between the insistence on transparency and actual practices in the field.

3. Rise of Shadow AI

The prevalence of shadow AI, unapproved AI usage within organizations, has escalated sharply, with 76% of organizations recognizing it as a significant issue—up from 61% the previous year. However, only 34% are leveraging external partnerships for AI threat detection, suggesting a gap where awareness is outstripping governance capabilities.

4. Misalignment in Ownership and Funding

Internal responsibility for AI security remains blurry, with 73% of organizations reporting conflicts over ownership of these controls. While there is a clear acknowledgment of the risks associated with AI, budget allocations reveal a gap, with over 40% of organizations investing less than 10% of their security budget into AI security.

The Evolution of AI Threats

The report highlights that the evolution of AI from basic applications to agentic systems capable of self-direction requires a reevaluation of existing security measures. Marta Janus, Principal Security Researcher, points out that simple model flaws can escalate to significant operational vulnerabilities, fundamentally changing the traditional threat model. The rapid growth of agentic systems, capable of carrying out tasks autonomously, makes them a prime target for cyberattacks.

New Trends Shaping the Threat Landscape

The landscape for AI threats is evolving due to three key shifts:

• - Increased Agentic Action: Agentic systems are moving quickly from experimental phases to widespread deployment.
• - Mainstream Autonomous Reasoning: Self-improving models leading to broader system influences enhance both accuracy and the damage potential if compromised.
• - Decentralization of AI Models: Smaller AI models deployed on various devices open up new security vulnerabilities, especially in regulated sectors.

Conclusion

The findings in HiddenLayer's 2026 AI Threat Landscape Report paint a complex picture of the current state of AI security and the significant challenges enterprises face as they navigate the operational integration of agentic AI. Their AI Security Platform offers vital protection mechanisms throughout the AI lifecycle, ensuring organizations fortify their defenses against evolving threats.

Access the complete report here.

About HiddenLayer

HiddenLayer provides comprehensive security for agentic, generative, and predictive AI applications, from asset discovery to runtime protection. With a focus on protecting intellectual property and supporting regulatory compliance, HiddenLayer equips organizations to confidently adopt and scale AI initiatives.