Amazon Bedrock Adds Web Search So OpenAI Models Stop Guessing

Amazon has made Web Search generally available on Amazon Bedrock, a built-in tool that lets OpenAI's GPT-5.4, GPT-5.5 and GPT-5.6 models, including the Sol, Terra and Luna variants, pull in current web information without leaving AWS's own infrastructure.

The pitch is grounding without the usual integration tax. Before this, giving a hosted model access to live web data meant wiring up a third-party search API, handling a separate vendor relationship, and running that vendor through your own security review. Web Search on Bedrock reduces that to a single parameter on an existing API call. According to AWS's announcement, it is built entirely by Amazon, drawing on a web index spanning tens of billions of documents and a knowledge graph for verified facts, and it keeps data inside a secured AWS environment with zero data egress to an outside search provider.

That last point is the one enterprise buyers will care about most. A lot of regulated industries cannot use a model with open web access if that access routes queries through an external search vendor's infrastructure. Keeping the entire search and grounding step inside AWS's own boundary removes an entire category of vendor risk assessment that would otherwise sit between a team and a working prototype.

The feature is live today in US East (N. Virginia), US East (Ohio) and US West (Oregon), with wider regional availability presumably to follow as it usually does with Bedrock features.

This is a smaller announcement than a new model, but it is the kind of infrastructure move that quietly decides which platform a cautious enterprise team actually builds on. Model quality gets the headlines. Procurement and security review are what determine whether a team is allowed to use that model on real client data at all. A feature that turns grounding a model in current information from a six-week vendor evaluation into a single API parameter is the sort of thing that does not trend on launch day and then shows up, a year later, as the quiet reason one cloud platform won a particular enterprise contract over another.