AI News Roundup – MIT report finds 95% of corporate generative AI projects are failing, AI moderators for social media fall short of expectations, AI chatbot deployed on China’s space station, and more

To help you stay on top of the latest news, our AI practice group has compiled a roundup of the developments we are following.

• • Fortune reports on a new publication from the Massachusetts Institute of Technology (MIT) that claims that 95% of AI projects introduced at companies fall short of success. The report, “The GenAI Divide: State of AI in Business 2025,” was prepared by MIT’s NANDA initiative, and found that only about 5% of AI pilot programs achieve “rapid revenue acceleration,” and that the vast majority have little or no impact on corporate income. According to the lead author of the report, Aditya Challapally, the main obstacle to AI in corporate settings is not the quality of the AI models themselves, but poor integration of AI tools into business environments. Further, the report found that over half of generative AI budgets are generally allocated to sales and marketing tasks, rather than to the automation of back-office tasks where the return on investment is generally higher. The report also predicted that AI-caused workforce disruption will continue, but that due to unsanctioned use of AI by workers, further impacts of the technology may be difficult to measure. 
• • Bloomberg reports on the limitations of AI moderators on social media sites. Content moderation, the process of reviewing content on social media sites for compliance with platform rules and removal of non-compliant content, has been increasingly done with the assistance of AI tools, even as generative AI is used to create a growing amount of content on such sites. Several professional moderators, speaking with Bloomberg, said that AI is still not capable of identifying dangerous or harmful content at the same level as human ones, and that social media companies, in an effort to cut costs, are moving too quickly into a majority automated moderation system. Many social media sites, including Meta’s Facebook and Instagram and Elon Musk’s X, have had their moderation practices questioned, with critics citing the proliferation of violent and harmful content on the platforms in recent years while the companies slashed human content moderation jobs in favor of automated systems. 
• • WIRED reports on a new AI chatbot being deployed on China’s Tiangong space station. The chatbot, named Wukong after the monkey of Chinese legend, was introduced onto the space station last month. Though exact information on the model is limited, the model is claimed to be based on a Chinese open-source AI model, and then modified to focus on spaceflight data, navigation, and planning. One official in China’s space program told a state news agency that “this system can provide rapid and effective information support for complex operations and fault handling by crew members, improving work efficiency, in-orbit psychological support, and coordination between space and ground teams.” The chatbot has since supported the crew of Tiangong on a spacewalk mission, and acts as an intelligent assistant for the station’s taikonauts. 
• • The Verge reports on new features coming to Google’s Gemini AI systems, unveiled at a recent company event focused on the Pixel smartphone line. At the event, the company announced its latest series of Pixel devices, to be launched in late August, as well as additional features for Gemini Live, an AI assistant that can conduct real-time conversations with users and interact with items on a user’s device. With the new features, Gemini Live will be able to highlight items seen through the device’s camera; for instance, if a user is searching for a specific tool, the AI system will highlight the tool in the camera view to point it out to the user. Gemini Live will also be able to interact with other apps on user’s devices, such as messaging and telephone apps, and will receive an updated audio model to improve the text-to-speech capabilities of the AI system. 
• • The Financial Times reports that Meta will license image and video generation AI technology from Midjourney. Meta’s Chief AI Officer Alexandr Wang said on X that Meta would license Midjourney’s “aesthetic technology for our future models and products, bringing beauty to billions.” The move marks a shift in strategy for Meta, who had previously focused on in-house AI development, including its flagship Llama large language models, in part by offering nine-figure compensation to attract top AI talent. Meta also recently reorganized its AI group into four distinct teams, the fourth such reorganization in the past six months as the company’s AI offerings have been seen as lagging behind those of rivals Google and OpenAI. The company began using third-party tools for internal work earlier this year as well, though the company still remains dedicated to AI content on its social media platforms.