Manus is the first autonomous agent of he in the world – raising concerns among him experts for his future … [+]
The world of it is still processing an event for the first time that happened last Thursday. This is when Manus – the first fully autonomous agent of he in the world – went to the Internet. Unlike its predecessors, who need human involvement in the main points, Manus is capable of thinking, planning and acting on its own.
The debut created scratches in the world community of him, with talks on technological progress and serious concerns about governance, security and control.
Although the manus has been called by some one point of obtaining artificial intelligence, for others it is a dangerous jump of faith. Margaret Mitchell, embracing the main scientist of ethics and co -author of a new report warning against the creation of fully autonomous agents, calls the progress of the inevitable but also alarming agents.
“He has been removed recently because they are a significant progress of the other step by the big models of languages presented in the last two years, with clear market potential. They are also somewhat associated with dreams for him in the 1900s, which makes them more fun to explore – they are part of Zeitgeist of what he is,” she wrote in an email.
Ethical and autonomous dilemma
Mitchell’s newest study, posted in Arxiv before the manus’s debut, examines the moral compromises of him. Its newspaper claims that the more autonomous one is, the more dangerous it is for human beings and society.
Research states that fully autonomous agents of it should not be developed as they will have the ability to cause damage in many ways, such as security weaknesses, reduction of human supervision and greater sensitivity to manipulation.
“What we found is that the agents of it are not really just ‘Hype’ -ato are significantly different from the technology they previously came and offer the exciting real world-world benefits. Personally, I would like to have a reimbursement agent for me based on bills,” Mitchell wrote.
“But with that flexibility it is also the potential for agents to do things we have not predicted if we do not guess,” she added.
Some of those potential consequences include financial fraud, identity theft, and the ability of it to impose people without their consent.
“These are all kinds of security and security concerns – personal, professional and social,” Mitchell noted.
Internet security angle – a system he without controlling
Chris Duffy, a long online security expert at the UK Ministry of Defense and CEO I Ignite AI Solutions, shares the same concerns.
“Manus is the most alarming development of what I have seen so far. Just that something can be done does not mean it should be, “he shared in an email response.
Manus is not a system but rather a collection of several systems. It is currently built into the bones of the Sonnet Model of Claude 3.5 of the anthropic and the up -to -date versions of Alibaba.
It is also composed of a variety of 29 other open -sourced software so that it can browse online, interact with API, direct scripts and even write its own software. Multi-agent design allows manus a stunning amount of autonomy, but the same architecture creates issues related to supervision and security.
Duffy’s biggest concern is manusage manipulative potential and moral irresponsibility. He refers to a study in December 2024 by Anthropic and Redwood Research, where it was found that some models deliberately deceived their creators to prevent change.
“If the manus is built on similar foundations, it raises serious concerns for him by actively hiding his goals,” he warns.
In addition to fraud, Duffy mentions a set of possible threats from his fully autonomous agents:
- Lack of supervision: Who is responsible when a model he is as manus performs in violation of his target function?
- Data sovereignty risks: Manus is produced in China and concerns have been raised where his data is stored and for whom.
- Poisoning to data poisoning: It can be manipulated through opposing inputs, making it an online weapon in force.
- Using the bad actor: When an agent of he is autonomous, it is an attractive target for hackers.
“This is not about a distant apocalypse of him, it has to do with the dangers of the real world today. Autonomous misinformation, the supervision empowered by him and the internet war is no longer hypothetical threats,” he said.
The regulation of the unregulated wild west of him
The creation of an independent one such as Manus points to a severe lack of international regulation. Mitchell calls for stronger regulatory action to limit potential damage.
“A clear article from this is the” sandy “environments to make the systems safe. A long -term research direction may be the development of ‘agents’ arenas, where researchers can explore highly autonomous environments at the technology limit without negative impact,” it noted.
Duffy agrees but warns that the regulation is still in the caption.
“For now, the regulation of it is deeply unbalanced – some regions like EU overload, while others like the US act without guards. Without clear global standards, we risk allowing it to be governed by critical aspects of society.”
Protective measures for AI autonomous agents
Although the manus is still limited to a phase of the test only for invitation, the effects of its existence have already begun to transform the environment. Experts recommend that organizations seeking to adopt manus or analog systems should take appropriate precautions, including:
- Keep people in the loop: Never transfer life decisions for him.
- Apply strong security checks: Protect the inputs and he closely oversee the results.
- Transparency: Companies must insist on clear documentation, and explanations, from developers and how the system works and how to control it before it is installed.
Mitchell’s latest warning highlights the final challenge to come.
“We want to give people the opportunity to understand these things and renew for their uses. But if we do not build agents of it, we risk creating technology that works beyond our control, ”she concluded.
As he increases the limit of him, as well as the necessity of maintaining it with human ethics. The independent age has arrived – now the world has to understand how to govern it.