Researchers have found a way to address the problem of AI forgetting how to behave safely. The approach, developed by Saketh Bachu and his team, does not involve external filters or software patches, but rather changes how the model interprets dangerous inputs. The team tested their method on LLaVA 1.5, a vision language model, and aimed to ensure the model doesn't forget how to behave safely when it's been slimmed down.
Salt Security has announced the industry's first solution to secure API actions taken by AI agents. The company, which is showcasing its innovations at CrowdStrike Fal.Con 2025, has developed a solution that gives organizations visibility into every agent-driven action, governance to enforce the right posture, and real-time protection against AI agent abuse. The solution includes MCP Protect, which discovers and monitors all MCP servers and their interactions with AI agents, automatically assessing the risk of these interactions and protecting against malicious or unsafe usage. Additionally, Salt Security has introduced Agentic AI Governance, a new category of out-of-the-box security controls that enforce safe AI agent behavior.
Univers has been recognized as a leading company in the Gartner Magic Quadrant 2025 for global industrial IoT platforms. The company's application of AI enables businesses to recognize risks before they escalate, balance energy flows in their global operations, and accelerate the transition to carbon-efficient systems. Global integrators are adopting EnOS in large-scale implementations, while technology partners are expanding the AI capabilities to new applications in industry and energy. Univers plans to further expand the limits of AI for the energy sector, moving from monitoring and optimization to fully autonomous processes.
Univers, a leader in AI for energy, has been named the top-ranked company in Gartner's 2025 Magic Quadrant for Global Industrial IoT Platforms. The platform EnOS, developed by Univers, enables companies to resolve complex energy challenges with intelligent information based on data. With 365 million connected devices, 845 GW of renewable energy managed, and over 800 global clients, Univers offers a comprehensive solution for energy management, supporting companies throughout their energy transition.
Utah aims to be at the forefront of AI technology with an emphasis on safety, as leaders from academia, industry, and government collaborate to ensure the technology is both transformative and trustworthy. The Responsible AI Community Consortium at the University of Utah, led by Manish Parashar, a key figure in the state's AI initiative, seeks to attract top researchers and experts to the state. The University of Utah has invested $100 million in the Responsible AI Initiative, which aims to expand the state's cyberinfrastructure for safe testing of new AI models. The initiative also seeks to foster partnerships between government, industry, and academia to use AI to address real problems faced by Utah residents, such as improving air quality.
Researchers have discovered that routine eye tests may detect blood cancer early, thanks to groundbreaking research using artificial intelligence. A study analyzing over 1,300 UK patients found that microscopic changes in eye scans are linked to an increased risk of developing blood cancers, such as myeloma and leukaemia. Specifically, patients with these changes were seven times more likely to be diagnosed with myeloma and twice as likely to be diagnosed with leukaemia over the following decade.
The 2025 China Arbitration Summit will be held in Beijing with the support of over 70 international organizations. Experts and professionals in the arbitration community worldwide will gather to discuss cutting-edge topics in international arbitration, including the potential of AI and other digital intelligence technologies. The summit aims to improve the efficiency of arbitration, ensure equitable proceedings, and maintain the credibility of international arbitration. It will also strengthen exchanges and cooperation between China and Latin America in commercial international arbitration, promoting a more open, just, and sustainable dispute resolution system. The one-day event will provide simultaneous interpretation in Chinese and English, and will be held in a hybrid format with both online and offline participation.
The market for AI-powered smart cities is projected to reach USD 460 billion by 2025, growing at a CAGR of 28%. Cities in North America, Europe, and the Asia-Pacific region are implementing AI-powered traffic management, predictive policing, digital twins for climate resilience, and smart utilities. Intelligent transportation systems are utilizing real-time surveillance, curbside analytics, and autonomous mobility, driven by machine learning and computer vision. AI-driven traffic projects aim to slash commute times by over 20%, while pilot studies show reductions in urban congestion by up to 30%.
Alibaba has developed a new generation of foundation models that match the performance of larger predecessors while being significantly smaller and more cost-efficient. The new mechanism, unveiled by Alibaba Cloud, has sparked excitement among AI experts who are increasingly concerned about the rising costs of scaling up models. This innovation is a significant breakthrough for Alibaba, one of the leading players in China's artificial intelligence boom. The new models are expected to herald "the future of efficient LLMs".