According to a report by the British Broadcasting Company on October 9th, American technology company Anthropic announced that it will prohibit users from insulting its AI in a 'continuous and unnecessary' manner, which has sparked attention and controversy.
Claude's developer Anthropic announced on the 8th that they are updating their usage policy to allow their tools to terminate interactions when users exhibit "cruel" behavior. The company states that this has been done in "rare" circumstances before.
Now, such offensive behaviors will be included in a list of prohibited actions. This list also includes bullying others, encouraging self-harm, and creating private videos without consent.
Although Anthropic stated that the policy applies only to “extreme cases” of repeated insults, it once again sparked debates about how to interact with AI tools.
The company stated that this policy does not apply to “common user frustrations, refutations, dark-themed creations, or model testing and research.” However, it is still unclear what Anthropic considers to be “abuse or cruelty” towards their models.
The screenshot of the new policy has been widely shared and commented on social media. Some people praise it for “advocating courtesy”; however, others criticize the plan, calling it a “major mistake”. Some even suggest that this might weaken attention to the abuse of humans and animals.
This level of AI anthropomorphism is harmful. It makes people believe that AI has attributes that it does not actually possess.
Microsoft's AI head, Mustafa Suleiman, criticized competitor Anthropic in September for treating AI as human beings, and responded to its policy update with a sarcastic emoji of a 'melting face'.
This update also aims to address new risks in other areas, including the prohibition on using Claude for ‘deceptive activities’. Additionally, as the U.S. midterm elections approach, Claude must not be used to ‘deceive voters or disrupt elections’.
On the 8th, Anthropic also stated that its policies have previously prevented its tools from being used in the development of weapons.
But among all the updates, none have attracted as much attention and scrutiny as banning users from abusing chatbots. This also raises a controversial question: should we be rude or polite when using generative AI tools?
Some experts say that it is unnecessary to use words like "please" or "thank you" when writing or speaking with AI tools such as ChatGPT. Such usage also wastes linguistic resources—the large language models behind these generative AI tools need to process and respond using these resources.
Other experts argue that being rude with AI chatbots can have negative effects on human communication, as it could "influence the way we interact with others and treat them."
