Case Study
How to Secure the Use of AI at the Endpoint
A strategic guide on securing on-device AI workloads by defending the endpoint against emerging cyber risks. It outlines how to maintain an adversarial mindset and implement hardware-level protections to unlock AI innovation safely. Don't let cyber risk stall your AI transformation. Learn how to protect on-device AI workloads with a modern, hardware-first defense.
- Topic
- Security
- Published
- 4 May 2026
On-device AI is the next frontier of productivity, but it brings a new set of rules for cybersecurity. From model theft to data poisoning, the risks are real and the territory is largely unknown. To lead in the AI era, organizations must move beyond reactive security and adopt an adversarial mindset.
In this eBook, "How to Secure the Use of AI at the Endpoint," we provide a comprehensive framework for protecting your enterprise. Discover how a coordinated defense; combining Dell’s secure hardware with industry-leading security intelligence, enables your workforce to innovate with confidence.
Key Highlights:
The New Attack Surface: Understanding how on-device AI expands the threat landscape beyond traditional software vulnerabilities.
Model Integrity: Protecting the "brain" of the AI from being tampered with or stolen by malicious actors.
Data Poisoning Defense: Preventing attackers from introducing biased or malicious data into local training sets.
Prompt Injection Risks: Strategies to stop unauthorized users from manipulating AI outputs through malicious inputs.
Hardware-Level Security: Utilizing silicon-based protections (like Intel® vPro) to secure AI workloads at the foundational level.
The Adversarial Mindset: Why security teams must think like attackers to anticipate how AI tools might be exploited.
Zero Trust for AI: Applying the principle of "never trust, always verify" specifically to AI agents and their data access.
Supply Chain Vigilance: Ensuring the third-party AI models and libraries integrated into the fleet are verified and secure.
Real-Time Threat Detection: Using AI-powered security tools (CrowdStrike) to identify and stop attacks before they execute.
Persistence and Recovery: Ensuring "self-healing" capabilities (Absolute) to maintain security controls even if an endpoint is compromised.
Regulatory Compliance: Meeting the evolving legal requirements for data privacy and AI governance through localized processing.