America’s Federal Trade Commission (FTC) has been urged by the Center for AI and Digital Policy (CAIDP) to investigate OpenAI and freeze new GPT releases. The CAIDP is a non-profit research group focused on tech ethics in the domain of artificial intelligence. According to the organization, OpenAI has violated commerce laws by releasing GPT-4, its latest AI language model.
CAIDP claimed that the rollout of OpenAI’s GPT-powered text generation tools is “biased, deceptive, and a risk to public safety.”
The California-based startup, OpenAI, released GPT-4 earlier this month. The AI program has since gained a lot of attention for its ability to engage users in humanlike conversations, summarize lengthy documents, and even compose songs.
GPT-4 violates FTC guidelines, claims CAIDP
According to the Federal Trade Commission’s guidelines, the use of artificial intelligence must be “transparent, explainable, fair, and empirically sound while fostering accountability”. However, the CAIDP claims that the rollout of the fourth iteration of the GPT-4 model violates all these requirements.
The CAIDP filed a complaint with the FTC on Thursday, calling on the latter to investigate OpenAI’s GPT technology and regulate it if needed.
Section 5 of the FTC act prohibits deceptive and unfair acts and practices in commerce. However, OpenAI has violated this rule by misleading netizens through the rollout of a biased and deceptive product that poses a risk to public safety, the CAIDP wrote in the complaint.
In a report outlining GPT-4’s technical capabilities, OpenAI mentioned the model’s risks and shortcomings. The CAIDP highlighted these to make their point, urging the FTC to uphold its responsibility of investigating and prohibiting unfair and deceptive trade practices.
The group went on to ask the commission to put a halt on the commercial release of any more GPT products until the investigation is concluded and the risks mitigated. They called on the trade watchdog to independently assess OpenAI’s technology and make sure it’s compliant with the FTC rules before allowing GPT products back in the market.
OpenAI earlier admitted that GPT-4 is flawed
OpenAI has already acknowledged that the AI language model can generate harmful responses and perpetuate biases – something the CAIDP made sure to highlight. Furthermore, insecure back-end systems managing the company’s language models have raised concerns too.
Not too long ago, a bug in an open-source library leaked snippets of some users’ conversations and payment information to others.
Even the startup’s CEO, Sam Altman, admitted his fear about the potential misuse of AI technologies and their impact on society. He warned that criminals could launch cyberattacks and disinformation campaigns by leveraging the ability of language models to generate code, text, and images. Recently, Europol reported that ChatGPT is already being used for such nefarious purposes.
The FTC also published an advisory recently, warning companies against overhyping AI products.
This isn’t the first time that OpenAI has faced opposition over the release of its AI language. Concerns surrounding the potential misuse of such technologies and AI’s potential to promote bias have been around for a while now.
FTC’s advisory asked AI companies to check their claims exaggerating their capabilities and falsely claiming that the AI product can do a job better than non-AI products.
In its statements, the FTC also cautioned against the rollout of AI products that can be potentially misused for unethical purposes. However, cracking down on ChatGPT would be a huge step indeed, and it remains to be seen if the FTC will actually go ahead with it.
Read More
Georgianna Antes