The EU prohibits the use of virtual assistants supplied by AI during online meetings. These assistants are often used to transcribe, take notes or even save visuals and audio during a video.
In a Presentation of the European Commission Delivered European hubs of digital innovation earlier this month, there is a note on the “online label” slide which indicates that “no AI agent is authorized”.
AI agents are tools that can perform complex tasks and in several stages independently by interacting with applications, such as videoconferencing software. For example, Salesforce uses AI agents for The leaders of appeal sales.
The Commission confirmed that this presentation was the first time that this rule was imposed but refused to explain why when it was questioned by Politico. There is no specific EU legislation that covers AI agents, but AI models that will feed them will have to respect the strict and Controversial rules of the You have a document.
AI agents raise security problems
While notes for AI and other types of agent are not intrinsically a threat of security, according to a 2025 IA global expert reportSecurity risks come from the user who did not know what their AI agents do, from their innate ability to operate outside the user control and AI-AI potential interactions. These factors make AI agents less predictable than standard models.
SEE: How can you be used safely? Researchers from Harvard, MIT, IBM and Microsoft weigh
Technological companies must be cautious when promoting products that can reach an increasing amount without the awareness of the user. One of the biggest guard tales is that of Microsoft Recall, an AI tool that has enabled users to control their PC or search for files using natural language.
The convenience has had a cost: remember the captures captured of active windows every few seconds, saving them as chronology, raising concerns about confidentiality and use of data and leading to significant launch delays. Microsoft has since published a series of agents Specially designed to combat cyber players.
AI agents develop in prevalence
This did not prevent AI players from putting more control to their models. Anthropic added a Computer use function on his Chatbot Claude Sonnet In October 2024, which gave it the possibility to navigate the desktop applications, move the sliders, click on the buttons and type text. It is Deep research functionAnnounced this week, also responds to “agents” prompts, as does Microsoft's equivalent.
Last month, Openai has expanded its text and speech text tools to agent models, indicating their growing relevance. In January 2025, Operator OPENAI has announcedAn agent tool that runs the browser to do actions such as grocery command or booking visits independently.
SEE: The EU invests 1.3 billion euros to stimulate the adoption of AI and improve “digital skills”
Anthropic and Openai even work together to improve agent technology, the latter adding Management of the model context protocol of the oldAn open source standard to connect AI applications, including agents, to data standards. Anthropic has also united its forces with Databricks to help large corporate customers Build their own agents.
Techrepublic predicted at the end of 2024 that the The use of AI agents will increase this year. The CEO of Openai, Sam Altman, echoes this in a January blog articleBy saying: “We can see the first agents of the AI” join the workforce “and significantly change the production of companies.”
By 2028, 33% of corporate software applications will include an agency AI, against less than 1% in 2024, according to Gartner. A fifth of online stores interactions and at least 15% of daily work decisions will be carried out by agents that year.
