1-i.ai launches FALCON Verify to check AI answers inside ChatGPT and Claude

One Intelligence | 1-i.ai has launched FALCON Verify, a tool that проверates AI-generated responses inside ChatGPT and Claude and flags whether claims are supported, unsupported, contradicted or insufficiently evidenced. The release targets a growing reliability problem in generative AI as businesses and consumers use chatbots for research, writing and decisions.

Why it matters: - Generative AI can sound confident while mixing accurate details with incorrect numbers, unsupported claims, fabricated facts or bad conclusions. - 1-i.ai is positioning FALCON Verify as a way to show users which parts of an AI answer they can trust before they rely on it. - The tool is aimed at workflows where mistakes can create real costs, including research, analysis, sales, customer support and business decisions.

What happened: - One Intelligence | 1-i.ai announced the availability of FALCON Verify on Sept. 23, 2026. - FALCON Verify works inside ChatGPT and Claude. - Users can connect it in under a minute. - The free plan includes 25 checks per month. - The product is available now.

The details: - FALCON Verify checks AI-generated responses statement by statement against available evidence. - The system returns four verdicts: SUPPORTED, UNSUPPORTED, CONTRADICTED and INSUFFICIENT. - FALCON also shows what is wrong and what evidence drove the verdict. - Each verification creates a record with a shareable ID. - That record includes what was checked, what evidence was used and when the check happened. - Users trigger verification by saying, “FALCON, verify that.” - The workflow is designed as Ask → Answer → Verify → Decide. - FALCON Verify can be added as a connector to ChatGPT or Claude. - The product can also be integrated into software, AI agents and products. - Enterprise customers can deploy FALCON within their own network. - The free plan does not require a credit card. - The company said FALCON is built to make the basis for verification visible instead of asking users to trust another AI checking the first one.

Between the lines: - 1-i.ai is arguing that the core problem with AI is not just wrong answers, but answers that are partially right and therefore harder to audit. - The company is also trying to make verification part of the normal AI workflow, rather than a separate manual review step. - The release reflects a broader push toward AI governance and trust infrastructure as more organizations use large language models in production. - 1-i.ai cited outside research to make that case, including a 2025 EBU/BBC study that found 45% of more than 3,000 AI responses contained at least one significant issue. - The company also pointed to Stanford research showing 17% to 33% hallucination rates in specialized AI legal research tools. - In internal SEC filing tests, 54% of AI-generated statements were not supported by the provided filing. - In the example given, FALCON would separate a mix of correct and incorrect claims in a financial summary rather than judging the whole answer as right or wrong.

What's next: - 1-i.ai plans to extend FALCON Verify through integrations with software, AI agents and products. - Enterprise deployments inside customer networks suggest the company is targeting regulated or security-sensitive users. - FALCON Verify is part of 1-i.ai’s broader AI Control Plane, which covers orchestration, governance, observability, reliability and runtime operations. - The company said its broader mission is infrastructure to make AI robust, reliable and responsible. - More information is available through the company’s announcement, social updates on X and video channel.

The bottom line: - 1-i.ai is betting that AI adoption will increasingly depend on a second layer of verification, not just better prompts or more confident chatbots.