July 2025 has been a dynamic month in artificial intelligence. Advances in language modeling, business automation, regulation and tool accessibility have set the tone for these types of tools.
In the following article we present the main news that have emerged in the last month.
1. Release of GPT-4.5 and improvements in autonomous agents.
OpenAI has officially released GPT-4.5, an intermediate version between GPT-4 and the expected GPT-5. This new model is characterized by:
- Improved contextual understanding: GPT-4.5 shows markedly superior performance on long conversation tasks.
- More accurate autonomous agents: Improved execution of chained tasks, such as project planning, repetitive task automation and intelligent information search.
- Lower resource consumption: Despite its greater capacity, it requires less processing power, making it more accessible for hybrid cloud solutions.
This development is driving a new wave of productivity tools and personalized support, especially in areas such as customer service, project management, marketing and software development.
2. Google introduces Gemini 2 Ultra for businesses
Google DeepMind has introduced the second generation of its enterprise-oriented Gemini Ultra model. This version is designed to easily integrate with Google Workspace, allowing companies to automate:
- Drafting and review of documents.
- Predictive data analysis in Sheets.
- Workflow automation using Google Apps Script powered by AI.
Gemini 2 Ultra includes advanced visual capabilities, allowing you to work with images, videos and technical graphics. Its integration with Google Cloud Vertex AI makes it easy to create customized solutions without the need for advanced machine learning skills.
3. Revolution in generative video AI: Sora and its competitors
OpenAI has continued to develop Sora, its text-based video generative model. Although still in limited access, new examples have already been leaked with near-cinematic quality and significantly reduced rendering times.
Meanwhile, competitors such as Runway, Pika and Luma AI have introduced new versions of their models, with key functionalities such as:
- Video editing by voice instructions.
- Real-time artistic styling.
- Integration with videogame engines for cinematic generation.
This evolution is changing industries such as advertising, independent film, online training and content creation.
4. Europe accelerates regulation with the entry into force of the AI Law.
July was the month in which the effective implementation of the European Union’s Artificial Intelligence Act, passed in 2024, has begun. Companies developing or using AI systems in the European market must already:
- Categorize your systems according to risk level (minimum, limited, high or prohibited).
- Ensure the traceability of your algorithms.
- Report the use of generative AI and guarantee the authenticity of the content.
Furthermore, this regulation is expected to set a global standard, especially in Latin America and parts of Asia.
5. AI in the health sector: assisted diagnosis with generative AI
A team of researchers from MIT and the University of Zurich has published the results of a multicenter study testing a generative AI system trained on medical records and X-rays. This system was able to:
- Identify with 92% accuracy rare lung diseases.
- Suggest personalized treatments based on thousands of similar cases.
- Generate understandable explanations for physicians and patients.
The project, partially financed by the WHO, opens the door to a democratization of medical diagnosis in countries with few specialists, as well as to an improvement in the quality of service in public hospitals.
6. Advances in physical AI
Important developments have been presented in the field of robotics, particularly in the so-called physical AI: the application of generative and language models to the control of robots in real environments.
Boston Dynamics, now in collaboration with Nvidia, has shown advances in autonomous robots capable of understanding complex verbal commands, navigating industrial environments and collaborating with humans in warehouses.
For their part, universities such as Stanford and ETH Zurich have demonstrated robots capable of:
- Learn new tasks by watching videos.
- Correct your behavior in real time through natural human feedback.
- Adapt to changes in your environment without reprogramming.
This accelerates the move towards autonomous factories, advanced logistics and assistance in environments such as hospitals or nursing homes.
7. Emerging trends and short-term forecasts
Finally, July has confirmed some of the trends that will dominate the second half of 2025:
- Interconnected AI agents: Assistants will no longer be «silos» but will act as teams of collaborative agents.
- Digital identity and verification: The rise of generative content is driving new authenticity verification technologies, such as mandatory «digital bookmarks.»
- Personalized education at scale: Education platforms are using AI to tailor learning paths in real time for thousands of students.
July 2025 has been a month marked by the balance between innovation and regulation in the field of artificial intelligence. As models become more powerful and accessible, ethical, legal and transparency requirements are also growing. Companies that know how to leverage these tools responsibly will be better positioned.