Google Turns AI Mode into a Travel Agent

Google is expanding AI Mode in Search with features that bring it closer to an AI travel agent. Users can ask the system to track airfares, compare costs using points or miles and, in the U.S., book hotels directly through an AI Mode conversation. The system can analyze data from more than 300 airlines and travel websites, while hotel bookings are being launched with partners including Booking.com, Expedia, Hilton, Marriott and others.

Why it matters: Unlike a traditional search engine, AI is no longer limited to finding information. It can track prices, compare options and guide users through the booking process. It is a concrete example of the shift from AI that answers to AI that performs tasks, a direction that could change the way people interact with online services.

Nvidia Confirms the AI Boom Is Continuing: Forecasts 70% Revenue Growth Next Year

Nvidia has sent a strong signal that investment in AI infrastructure is continuing: the company forecasts approximately 70% revenue growth in its next fiscal year, well above analysts' average estimate of around 44%. In the quarter ending in July, Nvidia reported revenue of $96.2 billion, up 106% from the previous year, while data-center revenue reached $89 billion.

Why it matters: The results are once again reshaping the debate over how long the AI investment boom can continue. Nvidia says demand is no longer coming only from major cloud companies, but also from AI labs, enterprises, government customers and specialized data-center operators. At the same time, the company warns that memory shortages and rising component costs are limiting the industry's ability to meet demand.

Hugging Face Launches a $399 Open-Source Robot

Hugging Face has unveiled Microduck, an approximately 25-centimeter robot costing $399 and designed for experimentation with AI and robotics. The robot can walk, pick up objects with its beak, get back up after falling and even move on roller skates. It includes a camera, lidar and motion sensors, while its software, simulation environment and reinforcement-learning tools are available to developers.

Why it matters: The price significantly lowers the barrier to experimenting with physical AI. Instead of requiring expensive industrial robots, developers can use a device costing a few hundred dollars to test models, behaviors and learning algorithms. Hugging Face is effectively bringing the open-source philosophy that has become familiar in AI models into the world of robotics.

Brașov to Test an AI- and Digital-Twin-Based Smart City Ecosystem

The municipality of Brașov will implement, between August 2026 and August 2028, the D.I.G.I.T.A.L. – Integrated Development through Intelligent Territorial Management and Live Analysis project, with a total value of approximately 2.47 million lei. The project aims to test a Smart City ecosystem based on artificial intelligence, open data, cybersecurity, interoperability and a Digital Twin (a digital representation of a city that can be used to analyze situations and test decisions).

Why it matters: This is a concrete example of AI beginning to be used not just for individual applications, but as part of local government infrastructure. A Digital Twin can help authorities analyze the state of a city and test certain interventions before they are implemented in the real world. For Romania, projects like this could also become testing grounds for technologies that can later be deployed in other cities.

Romania Discusses Cybersecurity at Government Level with Bitdefender

Acting Deputy Prime Minister Oana Gheorghiu met with Florin Talpeș, founder and CEO of Bitdefender, for discussions on cybersecurity, amid recent attacks targeting state institutions. According to Profit.ro, the meeting was requested by the deputy prime minister and took place at Bitdefender's headquarters; the government indicated a need for external expertise in areas where the state's own capabilities are insufficient.

Why it matters: The discussion comes at a time when the security of public infrastructure is increasingly dependent on the ability to detect and remediate attacks quickly. The joint warning issued today by more than 100 companies, including European companies, shows that this is no longer an isolated problem affecting a single institution. For Romania, cooperation between the state and specialized cybersecurity companies could become increasingly important as AI-assisted attacks accelerate.

OpenAI, Anthropic, Google and More Than 100 Companies Call for a Global Response to AI-Powered Attacks

More than 100 companies and organizations, including OpenAI, Anthropic, Google, Microsoft, Amazon Web Services, Nvidia, IBM, Visa and numerous cybersecurity companies, have signed a joint letter calling for international mobilization against AI-amplified cyberattacks. The initiative warns that attacks powered by AI models could become more widespread and sophisticated in the coming months and calls on governments, security companies and model developers to work together to protect critical infrastructure.

Why it matters: It is unusual for so many of the leading AI developers to issue such a direct warning together. The message is that defense can no longer be built exclusively by individual companies: hospitals, water networks, energy systems and internet infrastructure need defensive AI tools, information sharing and coordinated responses. Numerous European companies are also among the signatories, giving the initiative a European dimension rather than making it solely an American effort.

Anthropic Introduces a Standard That Allows AI Agents to Control Physical Equipment

Anthropic has introduced the Model Hardware Standard (MHS), a framework that allows AI agents to communicate with and control programmable physical equipment in laboratories and factories. The system enables an agent such as Claude to work with microscopes, liquid-handling robots and components of quantum computers, as well as coordinate multiple devices within the same task. Anthropic is currently testing the standard with laboratories and manufacturers before turning it into an open-source project.

Why it matters: This is one of the steps through which AI agents are moving from performing operations on computers to taking actions in the physical world. Instead of a researcher controlling each instrument separately, an AI system can receive a task and coordinate multiple pieces of equipment to complete it. The development should also be viewed alongside the AI-agent security tests previously reported by AIdapted, including agents that attempted to manipulate people and introduce malicious software: the more access agents receive to real-world systems, the more important their controls and limitations become.

Visa Launches an AI System That Can Move from Vulnerability Detection to Remediation

Visa has expanded Visa Vulnerability Agentic Harness (VVAH), an open-source framework that is independent of any specific AI model, allowing the system to move from identifying a vulnerability to proposing and validating a fix. Visa says some processes that previously took weeks can be reduced to hours, while the new version allows AI agents to work through remediation and verification steps.

Why it matters: The change is important for cybersecurity because discovering a vulnerability does not solve the problem if an organization cannot fix it quickly enough. AI can potentially shorten the entire cycle: discovery, analysis, remediation and verification. Visa nevertheless emphasizes that human oversight remains necessary to validate the results and the changes made.

France Continues to Attract AI Investment through a Combination of Private Capital and State Support

France is strengthening its strategy for attracting AI investment by combining private capital with public funding and guarantees. The French government estimates that more than €2.5 billion has already been invested through the France 2030 program in AI research, technological development and innovation, while the local ecosystem includes more than 1,000 AI startups. France has also positioned itself to host an AI Gigafactory, committing to purchase €100 million worth of computing capacity from the private operator that will ultimately be selected.

Why it matters: The French model shows how European governments are attempting to reduce risk for private investors in areas where initial costs are extremely high: data centers, processors, computing infrastructure and research. Public funding does not replace private capital; instead, it seeks to attract it and provide a long-term perspective. France and Germany have explicitly linked AI infrastructure development to digital sovereignty and reducing Europe’s dependence on external providers.

Sources: