One Intelligence today announced the launch of FALCON Verify, a new AI verification capability designed to address a fundamental challenge with generative AI: an AI answer can sound convincing even when part, or all of it is wrong.
FALCON Verify is available directly inside ChatGPT and Claude, giving users a simple way to check important AI-generated answers before relying on them. Once connected, users can simply say, "FALCON, verify that." FALCON checks the AI response without requiring users to leave their AI conversation. The Free plan includes 25 checks per month.
AI assistants are increasingly used for research, analysis, writing, data analysis, sales, customer support and business decisions. Their responses can be fast, articulate and highly confident. But confidence is not evidence, and a confident answer is not necessarily a reliable one.
An AI response can combine accurate information with incorrect numbers, unsupported assertions, misinterpreted sources, fabricated details or conclusions that the evidence does not support, all presented with the same confidence and fluency.
That creates a fundamental trust problem: How do you know which parts of an AI answer you can actually rely on?
The problem spans AI systems and use cases. A 2025 EBU/ BBC study found 45% of more than 3,000 AI responses contained at least one significant issue. Stanford researchers found 17–33% hallucination rates among specialized AI legal research tools. In One Intelligence's internal SEC filing tests, 54% of AI-generated statements were not supported by the provided filing.
FALCON Verify makes verification part of the conversation in ChatGPT and Claude. When an answer matters, the user simply says: "FALCON, verify that."
FALCON checks the AI response against available evidence and shows what holds up, what does not, and why: SUPPORTED — The evidence supports what the AI said. UNSUPPORTED — The AI response is not backed by the available evidence. CONTRADICTED — The evidence says something different than AI response. INSUFFICIENT — The available evidence cannot confirm the AI response.
FALCON is designed to make the basis for verification visible rather than asking users to blindly trust another AI checking the first one. Every verification also creates a record with a shareable ID, providing a trace of what was checked, against what evidence, and when.
An AI answer does not have to be completely wrong to lead someone to the wrong conclusion. For example, an AI might report that revenue grew 18% to $2.4 billion, operating margin was 21%, and guidance was raised. If the source actually shows 12% growth, a 21% margin and reaffirmed guidance, FALCON checks the statements individually, identifying what is supported and what is contradicted, along with the evidence behind each verdict.
The challenge with AI is no longer simply getting an answer. AI can produce an impressive answer in seconds. The harder question is whether you should rely on it. By bringing FALCON Verify directly into ChatGPT and Claude, verification becomes part of the normal AI experience.
Infrastructure for More Trustworthy AI
FALCON Verify is part of 1-i.ai's broader infrastructure for robust, reliable, and responsible AI. The 1-i.ai AI Control Plane helps organizations build, govern and operate AI across applications, models and providers. FALCON Verify extends that infrastructure to the AI answer.
About One Intelligence
One Intelligence builds infrastructure for production AI. Its AI Control Plane brings together orchestration, governance, observability, reliability and runtime operations across AI applications, models, providers and enterprise systems.
