# AI News Digest - 2026-01-12

1. Google removed AI-generated overview results for certain medical queries after an investigation found they had provided misleading and potentially dangerous information, including incorrect dietary advice for pancreatic cancer.

2. OpenAI acquired the founding team of executive-coaching startup Convogo to bolster its AI cloud and product development and to integrate the team's expertise into its infrastructure and offerings.

3. Lightricks open-sourced LTX-2, a 19-billion-parameter text-to-audio-video generation model that produced synchronized audio and video from text prompts and was reported to be faster than competing systems.

4. Chinese researchers developed UniCorn, a framework that diagnosed aphasia-like disorders in multimodal image models and enabled models to identify and repair their own weaknesses.

5. Princeton University, UCLA, and University of Pennsylvania researchers proposed web world models that used standard web code to create persistent, rule-defined environments and used language models to populate those environments for consistent agent exploration.

# References

1. [https://www.theverge.com/news/860356/google-pulls-alarming-dangerous-medical-ai-overviews](https://www.theverge.com/news/860356/google-pulls-alarming-dangerous-medical-ai-overviews)

[Google pulls AI overviews for some medical searches](https://www.theverge.com/news/860356/google-pulls-alarming-dangerous-medical-ai-overviews)

1. [https://the-decoder.com/convogos-founders-join-openai-to-close-the-gap-between-ai-potential-and-actual-use/](https://the-decoder.com/convogos-founders-join-openai-to-close-the-gap-between-ai-potential-and-actual-use/)

[Convogo's founders join OpenAI to close the gap between AI potential and actual use](https://the-decoder.com/convogos-founders-join-openai-to-close-the-gap-between-ai-potential-and-actual-use/)

1. [https://the-decoder.com/lightricks-open-sources-ai-video-model-ltx-2-challenges-sora-and-veo/](https://the-decoder.com/lightricks-open-sources-ai-video-model-ltx-2-challenges-sora-and-veo/)

[Lightricks open-sources AI video model LTX-2, challenges Sora and Veo](https://the-decoder.com/lightricks-open-sources-ai-video-model-ltx-2-challenges-sora-and-veo/)

1. [https://the-decoder.com/chinese-researchers-diagnose-ai-image-models-with-aphasia-like-disorder-develop-self-healing-framework/](https://the-decoder.com/chinese-researchers-diagnose-ai-image-models-with-aphasia-like-disorder-develop-self-healing-framework/)

[Chinese researchers diagnose AI image models with aphasia-like disorder, develop self-healing framework](https://the-decoder.com/chinese-researchers-diagnose-ai-image-models-with-aphasia-like-disorder-develop-self-healing-framework/)

1. [https://the-decoder.com/web-world-models-could-give-ai-agents-consistent-environments-to-explore/](https://the-decoder.com/web-world-models-could-give-ai-agents-consistent-environments-to-explore/)

[Web world models could give AI agents consistent environments to explore](https://the-decoder.com/web-world-models-could-give-ai-agents-consistent-environments-to-explore/)

For the site tree, see the [root Markdown](https://ixtj.dev/.md).