EU prepares “EU Kids Act”: restrictions for children and “safe by design” rules for digital services

The European Commission is preparing the “EU Kids Act”, a new legislative framework that would introduce age restrictions for social networks, video platforms, online games and AI chatbots. According to a Commission document reviewed by the European press, children under 15 would not be able to use certain services independently without parental consent, while platforms would have to introduce age-verification mechanisms and additional protection measures. The proposal explicitly includes chatbots and AI companions, considered services that may pose specific risks to minors. The Commission confirmed that the draft was due to be presented by Ursula von der Leyen and Henna Virkkunen on September 17, according to the European Commission.

Why it matters: The novelty is that Brussels is not treating an AI chatbot simply as another online service. The proposal also introduces the principle of “safe by design” (safety built into the product from the outset), meaning that platforms would have to prevent certain risks to children by design rather than waiting for parents to discover and subsequently disable problematic features. For AI companies, this could mean age verification, parental controls, usage limits and changes to the way conversations with minors are designed. The details of the proposal were presented by Euronews.

Bill Gates warns that governments are unprepared for AI and announces $1 billion for access to technology

Bill Gates warns that no government in the world is sufficiently prepared for the changes artificial intelligence will bring to society. In an interview with Reuters, the Microsoft co-founder pointed to risks involving jobs, cyberattacks and dependence on AI companions, while also highlighting the technology’s potential to improve people’s lives. At the same time, the Gates Foundation announced a $1 billion program over the next two years aimed at expanding access to AI in healthcare, education and agriculture and developing tools adapted to languages that are less well represented in current models, according to the Associated Press.

Why it matters: Gates introduces a dimension that is often missing from the competition between major AI companies: who actually benefits from the technology

ENISA warns that cyber defense must reach the speed of AI attacks

The EU cybersecurity agency ENISA warns that the development of frontier AI systems could turn cyberattacks into a race conducted at machine speed. In its report on cybersecurity in the age of frontier AI, the agency recommends developing operational capabilities able to respond to threats unfolding in minutes rather than days. ENISA warns that AI can accelerate vulnerability discovery, exploit development and the coordination of multiple stages of an attack, as the agency explains in its materials on AI and next-generation technologies.

Why it matters: The traditional security model assumes that defenders have time to analyze an alert, identify the vulnerability and then decide what action to take. If an AI agent can carry out these stages automatically, that advantage disappears. This is why ENISA refers to “machine-speed threats”: organizations need to be able to detect, assess and contain certain attacks almost automatically, while keeping human intervention for decisions with high impact.

Trump again dismisses warnings about AI dangers and faces pressure for rules

Donald Trump continues to dismiss warnings about the risks of accelerated AI development and opposes significant new federal restrictions. At the same time, political pressure for safeguards is growing in Congress, where it is no longer limited to Democrats: some Republicans are also calling for rules for the most powerful AI companies. The debate includes proposals for limits on advanced systems and mechanisms through which certain models could be shut down in dangerous situations, reports The Guardian.

Why it matters: The dispute is beginning to move beyond the technology world and become a major political issue in the United States. On one side, the Trump administration emphasizes competitiveness and maintaining America’s lead over China; on the other, warnings from industry leaders are adding to pressure for regulation. If this divide becomes an electoral issue, the way the US regulates AI could become one of the major political themes of the coming years. See also Aidapted articles: “Calls grow for slowing frontier AI development, but no one wants to be first to hit the brakes” and “The Great Debate Over AI RISK: Which Side Are the People Who Matter On?”

China raises the stakes: Claude and GPT-5.5-Cyber presented as threats to infrastructure

China is moving from general warnings about AI to explicitly naming specific American models. Chen Yixin, China’s Minister of State Security, mentioned Claude Mythos, developed by Anthropic, and GPT-5.5-Cyber, developed by OpenAI, in an article published in China Cyberspace magazine. The official claims that such models increase the efficiency of vulnerability discovery and malware development and could amplify risks to Chinese infrastructure. The information was reported by Aidapted in an article published today “China accuses the US of technological hypocrisy while building its own barriers”.

Why it matters: The fact that Beijing is naming Western commercial models in an official security document changes the nature of the dispute. Claude and GPT are no longer viewed solely as competing technology products, but also as tools with strategic and military implications. The development could fuel fragmentation of the AI ecosystem: the US, China and Europe could end up developing not only different rules, but also increasingly separate infrastructures, models and security systems.

Apple launches iOS 27, and the new Siri AI enters the iPhone ecosystem

Apple began distributing iOS 27 on September 14, and one of the most important new features is Siri AI, the next generation of Apple’s assistant based on Apple Intelligence. In beta, Siri can understand the user’s personal context, search for information in photos, emails and notes, and perform actions across apps. Apple points out, however, that Siri AI is not initially available in the European Union on iOS, and access is limited to compatible devices and languages, according to Apple.

Why it matters: This marks the shift from a chatbot that users open separately to an AI that becomes part of the operating system. Siri does not merely have to answer questions; it can use personal context and take actions within the user’s apps. For Europe, however, there is a separate issue: Apple had already confirmed that Siri AI would not initially be available on iPhone and iPad in the EU, creating an important difference between the AI experience offered to European users and that offered in other markets.

Anthropic launches Claude for Financial Advisors and connects it to financial firms’ infrastructure

Anthropic has launched Claude for Financial Advisors, a service that allows financial advisors to connect Claude to the custody, portfolio management, CRM and planning platforms they already use. The company provides connectors and workflows for research, meeting preparation and documentation, and the solution can be integrated with tools used in the financial industry. Anthropic recommends the Enterprise version for financial advisory firms because it includes audit logs required to keep records of activities, according to Anthropic.

Why it matters: Claude is beginning to be connected to the systems where a company’s real data and processes reside. That is precisely why requirements such as audit logs are emerging: when AI participates in professional financial activities, what matters is not only the answer it produces, but also what data it accessed, what actions it performed and who can subsequently verify the process.

Exein raises $270 million for “Physical AI” security

Italian company Exein has announced $270 million in funding, at a valuation of $1.7 billion, to develop a security infrastructure dedicated to “Physical AI” systems — robots, drones, vehicles and other machines that use AI to make decisions and act in the real world. The round was led by Headline and included Sofina, Goldman Sachs, EIB Group through the European Tech Champions Initiative, KfW Capital and existing investors. Exein says it monitors approximately two billion devices and observes around 5,000 new, non-repeated attacks per week across its network, according to the company.

Why it matters: “Physical AI” changes the stakes of security. If a compromised AI operates a chatbot, the consequence may be misinformation; if the same type of system controls a vehicle, an industrial robot or infrastructure, a vulnerability can produce physical consequences. Exein is attempting to build what it calls a “security layer” for Physical AI.