Amazon unveils new AI agents, trying to thread the needle between autonomy and human control
Amazon Web Services is announcing a new set of AI agents for businesses, developers, and individual users, capable of doing everything from fixing security vulnerabilities to intercepting email.
Agents, unveiled at the AWS Conference in New York, represents an effort to increase autonomy while ultimately keeping humans in control of how much AI does on its own. It’s part of a broader effort by Amazon and others to make AI powerful without letting it run.
The new security agent, called AWS Continuum, starts in a supervised “learning mode” and gains the right to act on its own only as customers give permission, step by step.
Amazon Quick AI Assistant will now allow users to create their own back-end agents in a simple language to handle tasks such as tracking business deals or flagging regulatory changes.
Amazon quickly offered a redesigned task feed that checks email, messages, and calendar items in one important view; new links to services including Adobe, Figma, Snowflake, and WhatsApp; and the ability to tap multiple connected services to answer a single question.
On the developer side, AWS also pushes its coding agents to take on more of the grunt work, checking and testing new code before it’s deployed and cleaning up old code, while leaving the final decision to compile or deploy in human hands. A new iPhone app for Kiro, the company’s AI coding assistant, will let developers start and monitor that work from their phones.
Deepak Singh, the AWS VP who leads the Kiro team, said the big idea is to take the back-end work that AI has done away from humans — reviewing code, testing security findings, keeping software up-to-date — and letting agents handle it with minimal human intervention.
The faster AI writes code and uncovers problems, he said, the more people there are to review, test, and maintain: “Those are all nice problems to have, but they’re real problems.”
AWS also expanded AgentCore, its platform for building agents, and introduced AWS Context, a service that organizes enterprise data so agents can think about it.
Announcing the new Continuum security agent, AWS revealed a proliferation of powerful AI models – notably the Anthropic Claude Mythos – that can now detect software flaws and synthesize them into critical attacks faster than any human team can respond.
Amazon made headlines for raising concerns about those same models, reportedly warning the Trump administration of security risks in Anthropic’s more advanced AI, before a government order forced the lab to take its two new models offline.
Continuum starts with code vulnerabilities, and AWS says it will expand to other security features in the future. It works on problems the way a team of people would, given the time: evaluate the findings, assess whether the vulnerability is exploitable, and then propose a fix, with an estimate of what another change might break.
In cases where the customer has given the agent autonomy, Continuum can implement the configuration itself, incorporating the transition into the existing supply pipeline.
Neha Rungta, AWS director of applied science, said in an interview that this kind of speed is necessary given the rapidity of threats. AI can now group small errors together, he said, combining two findings of medium and low severity into something serious.
“That was something that would have taken a lot of effort, expertise, and determination for an attacker to get through – so the floor has been lowered,” said Rungta, who led the operation at Continuum. “The goal is to raise that floor again.”
