New research shows AI behavior can be less safe with robots - NBC News

NBC News has published a video segment examining research on the safety of artificial intelligence and robots. The segment is titled 'New research shows AI behavior can be less safe with robots' and appears in an NBC News best-of playlist. In the segment, NBC News correspondent Gadi Schwartz interviews Jay Choi, chief executive of Robocurve, and Edward Sun, a member of technical staff at Robocurve, about that research.

According to the segment's description, the discussion covers tests that included models from both OpenAI and Anthropic. Those two companies are the only model providers identified by name in the accompanying text. The excerpt does not name specific model versions, describe how the tests were constructed, or state a numerical result. It also does not give a date for the research or for the tests.

The segment frames the topic as research into whether AI systems behave less safely when used with robots, and presents Robocurve as the organisation carrying out that work. Choi is identified by his title as the company's chief executive, and Sun as a member of its technical staff. OpenAI and Anthropic are cited in connection with tests that the segment says included their models. The excerpt supplies no figures, no dates and no description of how the tests were run, and it does not state whether the research has been peer reviewed or published in written form.