CVE-2026-108960 Tiiny-AI CVE debrief

Who should care

Defenders responsible for Tiiny-AI PowerInfer deployments should assess exposure and prioritize verification and remediation. This includes reviewing the current version of PowerInfer, understanding the potential impact of the out-of-bounds read vulnerability, and taking steps to prevent exploitation. Security teams and vulnerability management teams should also be aware of this vulnerability and its potential impact on the organization.

Why it matters

Defenders should care about CVE-2026-108960 because it is a public exploit vulnerability in Tiiny-AI PowerInfer that can lead to out-of-bounds read, and they should verify affected versions and apply fixes to prevent exploitation.

• Verify affected versions and apply fixes to prevent exploitation
• Monitor for public exploits and assess exposure
• Assess the impact of a potential out-of-bounds read on the system

Technical summary

The vulnerability is in the create_striped_mat_to_gpu function of the ggml-cuda.cu file in the FFN GPU-split Loader component of PowerInfer. This function has an out-of-bounds read vulnerability. The exploit is public and may be used. Defenders should prioritize verifying the affected version and applying the fix from the vendor. The vulnerability allows for an out-of-bounds read, which could lead to information disclosure or system crashes. Affected product deployments should be identified and verified for exposure.

Defensive priority

Defenders should prioritize verifying the affected version and applying the fix from the vendor, as the exploit is public.

Recommended defensive actions

• Verify the version of PowerInfer and check if it is affected
• Apply the fix from the vendor
• Monitor for public exploits
• Review compensating controls for exposed systems while remediation is scheduled and verified
• Check relevant monitoring, detection, and logs for exposed assets that need extra review
• Track exceptions, retest remediated assets, and close the item only after evidence is documented
• Confirm whether affected product deployments exist in managed environments and assign an owner for follow-up

Evidence notes

The CVE record and source item provide details about the vulnerability in Tiiny-AI PowerInfer, specifically in the create_striped_mat_to_gpu function of the ggml-cuda.cu file. The exploit is public and may be used. Defenders should verify affected versions and apply fixes to prevent exploitation. However, version information for affected or updated releases is not disclosed due to the rolling release system. Further verification tasks are needed to assess exposure and prioritize remediation.

Sources and references

Methodology and review provenance

AI-assisted synthesis based on stored public vulnerability evidence. System validation, approval state, and publication status do not by themselves establish human review of this revision. PatchSiren helps prioritize defensive review and does not prove exposure or remediation on any system.