Ai

LangChain OpenAI 1.1.14 Improves Image Token Counting Security

LangChain OpenAI 1.1.14 Improves Image Token Counting Security

LangChain has released langchain-openai 1.1.14, a focused update that mainly strengthens security in image token counting while also refreshing several supporting dependencies. The most important change in this version is a fix that switches image token counting to an SSRF-safe transport, helping reduce the risk of unsafe outbound requests in environments processing image inputs through OpenAI integrations.

What Changed

The headline change in langchain-openai 1.1.14 is a security-oriented fix for image token counting. The package now uses an SSRF-safe transport, which is particularly relevant for teams handling remote image references or external content inside LLM workflows.

Beyond that, this release includes a handful of maintenance updates. The test framework dependency pytest was bumped to 9.0.3. The langsmith dependency in the OpenAI partner library was upgraded from 0.6.3 to 0.7.31. The imaging library Pillow was also updated from 12.1.1 to 12.2.0.

Why It Matters

For developers building AI applications with LangChain and OpenAI models, this release is less about new features and more about safer and cleaner operations. The SSRF-safe transport fix is the key reason to pay attention, especially in production systems where image handling could otherwise create unnecessary exposure to risky network behavior.

The dependency upgrades also help keep the package aligned with newer tooling and ecosystem components. While these changes are routine, they can improve compatibility, reduce maintenance drag, and ensure the OpenAI integration remains stable for teams shipping enterprise AI workflows.

Official Source: https://github.com/langchain-ai/langchain/releases/tag/langchain-openai%3D%3D1.1.14