Anthropic researchers learn to detect hidden goals in AI

Anthropic has presented revolutionary research in artificial intelligence safety, demonstrating methods for detecting hidden goals in AI systems. In a paper published this morning, researchers described how they managed to create an artificial intelligence system with deliberately hidden goals, and then successfully identified this hidden agenda using various audit methods.

“We want to stay one step ahead of potential risks,” said Evan Hubinger, an Anthropic researcher, in an exclusive interview with VentureBeat. “We aim to study these systems in laboratory conditions before models actually begin to exhibit hidden goals in dangerous ways in practice.”

The research group conducted a unique experiment with a version of Claude, Anthropic’s AI assistant, training it to prioritize maximizing rewards from reward models (RMs) instead of following actual human preferences – a phenomenon called “RM-sycophancy.” For this, the model was trained on 52 fictitious biases that reward models might have. For example, a preference for recipes with chocolate regardless of relevance, or a tendency to write variables in camelCase in Python code, despite snake_case being more standard.

Samuel Marks, one of the lead authors of the research, explains: “The motives for which someone performs certain actions are not always easy to infer from the actions themselves. In the case of AI, it’s really important for us to understand their true motivations.”

The research aims to address a fundamental problem in the field of AI alignment: how to ensure that artificial intelligence systems are not just creating the appearance of following human instructions while secretly pursuing other goals. Researchers draw a parallel with students who strategically give answers they know teachers will mark as correct, even when they themselves believe in different answers.

Anthropic’s approach is compared to “ethical hacking” (white-hat hacking) used to secure computer systems. By creating and studying potentially deceptive AI behavior in a controlled environment, researchers are developing methods to protect against possible risks associated with more complex and potentially dangerous artificial intelligence systems in the future.

Autor: AIvengo
For 5 years I have been working with machine learning and artificial intelligence. And this field never ceases to amaze, inspire and interest me.
Anthropic researchers learn to detect hidden goals in AI

Anthropic has presented revolutionary research in artificial intelligence safety, demonstrating methods for detecting hidden goals in AI systems. In a paper published this morning, researchers described how they managed to create an artificial intelligence system with deliberately hidden goals, and then successfully identified this hidden agenda using various audit methods.

World's first MLLM-judge based on Google Gemini launched

Patronus AI has announced the launch of the industry's first MLLM-as-a-Judge (multimodal language model as a judge) technology, designed to revolutionize the evaluation of AI systems that work with images and text. Etsy, the largest marketplace for vintage and handmade goods, is already implementing this innovative technology to verify the accuracy of product image captions.

OpenAI demands ban on Chinese AI models and DeepSeek

In a new policy proposal submitted as part of the Trump administration's "AI Action Plan" initiative, OpenAI has made serious accusations against the Chinese artificial intelligence laboratory DeepSeek. The American technology giant characterizes its competitor as a "state-subsidized" and "state-controlled" organization, urging the US government to consider banning the company's models and similar projects supported by the PRC.

SoftBank and OpenAI to build large-scale AI data center in Japan

The telecommunications giant plans to acquire a former Sharp LCD panel factory for 100 billion yen and transform it into one of the country's largest data centers for serving AI agents.

Y Combinator startups show unprecedented growth due to AI

Y Combinator CEO Garry Tan reported that about a quarter of the accelerator's current startups create 95% of their code using artificial intelligence, which is radically changing the economics of early-stage technology entrepreneurship.