Top Highlights
- AI accelerators in devices can be exploited to perform privileged actions outside normal security controls, risking system compromise on over 100 million devices.
- Hardware vulnerabilities in chips like Qualcomm’s and AMD’s allow attackers to access sensitive data, take full control, or forge security attestations.
- The rise of AI-driven vulnerability discovery tools has led to real-world zero-day exploits, highlighting escalating attack sophistication and industry security gaps.
The Threat, Attack Techniques, and Targets
Security researchers warn that AI chips in connected devices are becoming a new cybersecurity risk. These chips are used in smartphones, industrial sensors, cars, and other connected devices. As companies work to put AI on the edge of networks, the risk grows.
Recent studies show that AI accelerators can be attacked. These chips handle machine learning tasks locally in devices. Researchers found that six out of seven tested chips from big vendors could be manipulated. Attackers can make these chips perform privileged actions for malicious apps. This affects more than 128 chip designs and over 100 million devices.
One type of attack called “confused deputy attack” is common. In these attacks, AI chips do more than normal because they are outside many traditional security controls. Additionally, hardware flaws have been found in Qualcomm Snapdragon chips, AMD server processors, and several device models. The flaws can allow attackers to access data, control sensors, or even take full control of devices.
Targets include smartphones, industrial systems, cloud servers, and automotive components. The use of AI hardware at the edge increases, but so do the attack surfaces. Attackers are also using AI itself to find new vulnerabilities and develop exploits quickly.
Impact, Security Implications, and Remediation Guidance
The impact of these threats is significant. Attackers may gain unauthorized access to sensitive data, control devices, or disrupt operations. The vulnerabilities could lead to data breaches, device takeovers, or even wider network compromises.
Security implications include the need for stronger protections for AI hardware. Many organizations focus on software security but overlook hardware risks. As edge AI deployment grows, so does the chance of malicious exploitation. The recent hardware flaws highlight the urgent need for better security standards for AI chips.
If you suspect or discover vulnerabilities, it is important to follow guidance from the device or chip vendor. Manufacturers like Qualcomm and AMD have released patches to fix some issues. For other devices, consult the relevant vendor or authority for remediation steps. Staying updated with firmware and security patches is essential to reduce risk.
Continue Your Tech Journey
Learn how the Internet of Things (IoT) is transforming everyday life.
Discover archived knowledge and digital history on the Internet Archive.
ThreatIntel-V1
