CVE-2026-7482: CVE-2026-7482: Ollama before 0.17.1 contains a heap out-of-bounds read vulnerability in the GGUF model lo…

What it is

Ollama before 0.17.1 contains a heap out-of-bounds read vulnerability in the GGUF model loader. The /api/create endpoint accepts an attacker-supplied GGUF file in which the declared tensor offset and size exceed the file's actual length; during quantization in fs/ggml/gguf.go and server/quantization.go (WriteTo()), the server reads past the allocated heap buffer. The leaked memory contents may include environment variables, API keys, system prompts, and concurrent users' conversation data, and can be exfiltrated by uploading the resulting model artifact through the /api/push endpoint to an attacker-controlled registry. The /api/create and /api/push endpoints have no authentication in the upstream distribution. Default deployments bind to 127.0.0.1, but the documented OLLAMA_HOST=0.0.0.0 configuration is widely used in practice (large public-internet exposure observed).

Exploitation evidence

First exploit evidence seen: 2026-05-05 (earliest date across the sources below).

Exploited in the wild

No source on this site lists in-the-wild exploitation yet (CISA KEV, VulnCheck KEV).

Public exploit code

GitHub proof-of-concept repositories

3 public repositories from 3 accounts, first seen 2026-05-05. Most starred: 0x0OZ/CVE-2026-7482-PoC (6 stars), view repository.

Verify before running: public PoCs can be malicious. Read the code and only run it in an isolated lab. A PoC means exploit code is public, not that this CVE is exploited in the wild.

Sources: nomi-sec/PoC-in-GitHub.

How to fix it

Fixed versions

• ollama ollama: upgrade to 0.17.1 or later

Derived from the NVD CPE configuration (versionEndExcluding / versionEndIncluding). Always confirm with the vendor advisory.

Vendor advisories and patches

• https://github.com/ollama/ollama/commit/88d57d0483cca907e0b23a968c83627a20b21047 (Patch)
• https://github.com/ollama/ollama/pull/14406 (Issue Tracking, Patch)
• https://github.com/ollama/ollama/releases/tag/v0.17.1 (Release Notes)

MITRE ATT&CK mitigations for this chain

• M1050 Exploit Protection (for T1203) : Security applications that look for behavior used during exploitation such as Windows Defender Exploit Guard (WDEG) and the Enhanced Mitigation Experience Toolkit (EMET) can be used to mitigate some exploitation behavior. Control flow integrity checking is another way to potentially identify and stop a software exploi
• M1051 Update Software (for T1203) : Perform regular software updates to mitigate exploitation risk. Keeping software up-to-date with the latest security patches helps prevent adversaries from exploiting known vulnerabilities in client software, reducing the risk of successful attacks.
• M1048 Application Isolation and Sandboxing (for T1203) : Browser sandboxes can be used to mitigate some of the impact of exploitation, but sandbox escapes may still exist. Other types of virtualization and application microsegmentation may also mitigate the impact of client-side exploitation. Risks of additional exploits and weaknesses in those systems may still exist.

Detection Generic guidance

• Execution: Watch for the service process spawning shells or script interpreters (cmd.exe, powershell, sh, bash, python) via EDR process-tree rules.

Response checklist Generic guidance

1. Inventory: find every instance of the affected product and version (include internet-facing and forgotten ones).
2. Patch or mitigate per the vendor advisory; if you cannot patch, remove internet exposure or apply the vendor workaround.
3. Assume compromise if the system was exposed before patching: hunt for the indicators in the advisory and review logs back to the disclosure date.
4. Rotate credentials and secrets stored on or used by the affected system if compromise cannot be ruled out.