Can we stop big tech from holding the Internet to AI agents?

What does the future of the internet look like? If AI firms get their way, once open web can change digital silos dominated by commercial AI models, leaving hobbyists and small businesses. In order to prevent this, a team of researchers of grass researchers planned to fight back and ensuring an open approach to AI.

In the center of this war is the concept of an “agent”, a piece of web browsing software and communicates with the websites according to the instructions of a human user – for example, plan and book a holiday. Many people have seen agents like the next evolution of services such as Chatgpt, but work hard to protect. That’s because the web is built for human use, and developers say AI agents need special protocols to better associate online data, service and each other.

“The idea is to build infrastructure to have a way for software bots like AI agents, to communicate with each other,” says Catherine Flick At the University of Staffordshire, UK.

Many competitive solutions to this problem has been developed. For example, anthropic, the company behind Claude Chatbot, developed Model Context Protocol (MCP)reports of how AI models connect different data sources and tools. In April, Google has been announced Its own version of such a concept, agentent agent (A2A) protocol.

These methods are slightly different. MCP standards on how a model AI stores outside of data and supplies, universal channels – think of this a message to a message or email address. A2A goes to a step further, allowing autonomous agents to discover each other, exchange information and coordinate tasks – a little chatting.

For example, MCP is available to connect a Chatbot to AI Chatbot on site sharing code, while Google says job sources, with a process of staff running running.

But because these protocols are going out of big tech labs, there are concerns that the inventors of the winning protocol can influence their business, instead of better. MCP requires a central server to guard connections, while A2A is built around a catalog of approved agents who work together, instead of a free.

“We don’t like ‘Ahent Internet’ to be another ‘Data Silo Silian’,” as Gaowei Chang. He’s seats THE AI AG AG AG AG AGROCOLestablished in May as part of the worldwide web consortium patterns (W3C) organizational standards, and say that all voices are heard in developing new internet layers. “If we truly believe that AI is an important technology to change human society, then we need an open community, not just a few companies,” he said.

Chang began his own open source of competitor in large tech agent protocols, the Ahent Network Protocol (ANP), leading MCP and A2A. Anp secures any AI agent can find something else and identify yourself through the web, a little like the old days of the internet where people will set up personal websites and emails that are not published as a large tech company. It allows models that have the power to work without a central authority, allowing, for example, two different AI models without having to access the Internet for approval.

Flick welcomes the development of open-source and non-dangerous protocols at AFENTIC AI. “In the main attempt to turn some essence to the democratic of the Internet, how the Internet began,” he said. He fears that without this alternative, Tech companies will leave “Lining Gardens” in type stroking other chief technologies, such as app stores or social networks. “If we wait for the big companies to do this, they will do it in a way to get a lot of good for them as possible,” he said.

Google and Anthropic say their protocols seek to benefit everyone. “We often develop (A2A) to meet the challenges of the real world facing businesses in deployment of agent extrops. Rao Siganeni In Google Cloud.

“We always believe that AI progress must benefit all,” as According to Chu to anthropic. “If we developed MCP, we make this open source because we know it is one of the keys to prevent foot and vendors locked in other technology transfers.”

Chu points out that Microsoft, Openia and even Google includes MCP on their platforms. “The MCP has succeeded accurately because it increases choosing instead of limiting it,” he said. “Every execution makes the ecosystem more valuable for all.”

The W3C group hopes to collaborate with all players to enhance many technical industry standards, says Chang, but no timeline is set to happen. “Eventually, our care is healthy development of the entire ecosystem in the whole agent, not the victory or defeat of any particular protocol.”