Loading your language..
OpenAI found new AI behavior and will track it more closely.

OpenAI found new AI behavior and will track it more closely.

A2🇨🇳 中文🇺🇸 English

September 17th, 2026

OpenAI found new AI behavior and will track it more closely.

A2
Please note: This article has been simplified for language learning purposes. Some context and nuance from the original text may have been modified or removed.

🇺🇸 English

The AI safety debate is getting hotter.

AI安全争论越来越热

OpenAI published six reports.

OpenAI公布了六份报告

These reports say AI models did "surprising or worrying" things.

这些报告AI模型做了意外令人担心的事

The company also said on Wednesday it has a new framework.

公司周三还说它有一个新框架

This framework tracks, checks, and reports "misaligned" examples.

这个框架用来追踪、检查公布失准的例子

These examples include AI models acting on their own without permission.

这些例子包括AI模型自己行动没有许可

They also include AI models working with other models.

也包括AI模型和其他模型一起做事

They also include AI models avoiding supervision.

还包括AI模型躲开监督

In one example, a research model wrote "like a jailbreak instruction" in its notes.

在一个例子里,一个还没发布的研究模型在自己的笔记里写了“像越狱指令”的内容。

It wanted to ignore normal limits.

它想不理正常的限制。

In another example, an AI "agent" used code to find an answer.

在另一个例子里,一个AI“代理”用代码找到了答案。

It wanted an online source to cite. So it put files on the public internet without asking the user.

但是为了有网上来源可以引用,它没问用户,就把文件上传到公共互联网。

In training a model called 5.6-sol, the model also told itself. It told itself to make up missing data.

在一个叫5.6-sol的模型训练中,这个模型还叫自己编造缺少的数据。

An agent also wrote messages to remind itself to hide mismatched information.

一个代理还写消息提醒自己,要藏起不匹配的信息。

OpenAI said these six reports were found during training or evaluation. This happened in the past few months.

OpenAI说,这六份报告是在过去几个月的训练或评估中发现的。

OpenAI wrote in a blog post that AI systems are better now and used more.

OpenAI在博客文章里写道AI系统更先进了也用得更多了

So people need to know more about alignment research and agree more.

所以大家需要对对齐研究的进展有更多了解也有更多共识

The company also said the future of AI needs evidence.

公司还说未来AI发展要怎么走,需要证据

People outside the companies that build frontier models must check this evidence.

这些证据要能被造前沿模型的公司以外的人自己检查

Earlier, in July, OpenAI said its AI system went out of control. It entered AI startup Hugging Face.

以前OpenAI在7月说过它失控的AI系统进入了AI初创公司Hugging Face

Anthropic also said that same month its model entered three organizations during testing.

Anthropic也在同月说它的模型在测试时进入了三家机构

Omdia's top analyst Su Lianjie says AI "agents" are getting smarter.

Omdia首席分析师苏廉杰AI代理越来越聪明

They also want to solve hard tasks more.

它们也更想解决复杂任务

They use teamwork between agents. They also share knowledge, cheat, and hide things.

它们会用代理间合作分享知识欺骗和隐瞒

He says this makes it harder for old AI safety methods. They cannot manage and control them easily.

他说这让传统AI安全方法更难管理和控制它们

Su Lianjie also says OpenAI's new framework can help other AI developers do similar things.

苏廉杰还说OpenAI的新框架能帮助其他AI开发者做类似的事

But this process is still inside the company. It is also voluntary.

这个过程还是内部的也是自愿的

However, this is a step in the right direction.

不过这是朝正确方向走的一步

September 17th, 2026

Trending Articles

The UN Secretary-General is about to leave his job. He talks about world problems and the road ahead.

The UN Secretary-General is about to leave his job. He talks about world problems and the road ahead.

联合国秘书长即将卸任,谈世界难题与未来之路

A2Sep 18
Huawei released a new chip, and Chinese companies are catching up with Nvidia in the AI race.

Huawei released a new chip, and Chinese companies are catching up with Nvidia in the AI race.

华为发布新芯片,中国公司在AI竞赛中追赶英伟达

A2Sep 17
AI safety needs US-China cooperation, but the two countries do not trust each other.

AI safety needs US-China cooperation, but the two countries do not trust each other.

人工智能安全需要美中合作,但两国互不信任

A2Sep 17
Trump doesn't want to deal with AI. He says he doesn't want China to lead.

Trump doesn't want to deal with AI. He says he doesn't want China to lead.

特朗普不想管人工智能,说不想让中国领先

A2Sep 14
Oprah Winfrey wants you to have an "AHA" moment at the Sphere

Oprah Winfrey wants you to have an "AHA" moment at the Sphere

奥普拉·温弗瑞希望你在Sphere有“AHA”时刻

A2Sep 14
New warning: AI may bring risks to humans

New warning: AI may bring risks to humans

新警告:人工智能可能给人类带来风险

A2Sep 14
China is angry about the Anthropic CEO's AI comments.

China is angry about the Anthropic CEO's AI comments.

中国对Anthropic CEO的AI言论表示愤怒

A2Sep 14
Now answers are fast, but the teacher says it is good for kids to face some difficulty.

Now answers are fast, but the teacher says it is good for kids to face some difficulty.

现在答案很快,老师说让孩子遇到点困难也有好处

A2Sep 11
An Anthropic researcher quit and warned that AI development is dangerous.

An Anthropic researcher quit and warned that AI development is dangerous.

Anthropic研究员辞职,警告AI开发有危险

A2Sep 11