Update as of October 1, 2023: The browsing feature on GPT4 has been reinstated after thorough reviews and updates to ensure compliance with legal and ethical standards. Users will now notice enhanced safeguards to prevent access to restricted or paywalled content. The restoration of the browsing feature re-establishes GPT4’s ability to provide reliable and well-sourced information from the web, addressing earlier concerns regarding the generation of unsourced or fabricated content. Moreover, with the latest update, ChatGPT now boasts an “advanced data analysis” plugin for paid users. It allows you to perform more complex data analysis and interpretation tasks within the chat interface, such as analyzing trends, performing statistical tests, and generating descriptive statistics.
On July 3, 2023, OpenAI rolled back its “Browse with Bing” beta feature in the paid GPT-4 version of ChatGPT. Over a month later, the features remains disabled, with no indication of when, or if, it will be back. The decision was made following concerns that the feature was unintentionally displaying content that was not intended to be publicly accessible, including material behind paywalls. Specifically, the prompt “Print the Content of this Article” followed by the URL was being used to make such content accessible.
The Browse plugin allowed GPT4 to search the internet and provide users with information from a variety of sources. It was received with a lot of fanfare when it was launched because the chatbot could now produce reliable, well-written content. Without this plugin, GPT4 has returned to the dark ages of generating unreliable, unsourced — and sometimes, straight up fabricated — content.
This incident has once again brought back attention back to the murky landscape of legal and ethical issues that generative AI technologies face.The unfolding drama serves as a fitting entry point into the broader discourse around the legalities and policies encompassing AI, and the potential implications for businesses relying on these tools.
The Confusing Legal Implications
Generative AI and Large Language Models (LLMs) are nascent yet fast-developing technologies that open up a world of possibilities for individuals and businesses alike. But as is the norm with policy, by the time regulations are put in place for responsible usage of a technology, widespread adoption is already underway. For businesses, the legal ramifications can be enormous.
If businesses cannot trust the stability and legal security of AI tools, it could stymie AI adoption, undermining innovation and technological progress.
Take the OpenAI case, for instance. GPT4’s inadvertent capability violates the terms of service of many websites and infringes on copyright laws, putting both OpenAI and its users in a potentially precarious legal position. Rolling back the feature clears OpenAI. As it said, this is why it is a Beta version. In a tweet that OpenAI put out, it noted, “We are disabling Browse while we fix this — want to do right by content owners.” But, let’s remember that it is also a paid version. So while disabling the feature is doing right by the content owners, what about the paid users? This is a two-part problem.
First, how does the glitch affect past projects that may have unintentionally utilized content produced by GPT4 from sources that should not be publicly accessible? Imagine a content marketing campaign that is now littered with information from various sources that would ideally need to be paid for. Suddenly, it looks a lot more complicated and murky. Would the business be forced to take down the project? Or is it perhaps obligated to pay for the content retroactively? Both alternatives are unpalatable and translate into unforeseeable budget dents — sacrifice the project or cough up unallocated money. This is especially ironic because many companies turned to AI as a way to cut budgets in the first place.
This incident is not an anomaly but a symptom of the broader, murkier issues surrounding the use of generative AI and LLMs. Already, the very process of training these models is viewed by many as a form of copyright infringement, as they involve feeding the AI swaths of data of mixed copyright status. When the dust settles on this thorny debate around what constitutes infringement, organizations that have integrated AI into their business might be left dealing with unpredictable legal and operational ramifications. They could find themselves unwittingly in violation of copyright laws or privacy protections because of the actions of the AI systems they deploy.
Second, how does rolling back the feature affect future projects? From a business impact angle, unexpected feature changes, like the disabling of the browsing mode, can disrupt operations and potentially impact the bottom line. Many ChatGPT users believe that this glitch should have been detected and corrected at the Alpha stages and not at the Beta stage where businesses and individuals are paying to access the feature.
Ultimately, if businesses cannot trust the stability and legal security of AI tools, it could stymie AI adoption, undermining innovation and technological progress.
The Road Ahead: Treading with Caution
Given these complexities, what are the potential solutions? Clarity and regulation will be essential in this frontier of law and technology. Policymakers, legal experts, AI developers, and businesses need to collaborate to establish clear guidelines for the training and use of AI. These guidelines should protect intellectual property and personal privacy without penalizing subscribers or stifling the revolutionary potential that AI holds. Moreover, the AI development community needs to continue to enhance transparency and due diligence in their practices. If OpenAI’s recent predicament shows anything, it’s the importance of meticulous, ongoing oversight of AI tools.
As we continue to innovate and push the boundaries of what AI can do, it’s crucial to navigate these murky waters with both caution and foresight. In a game where the rules are still unclear, no party involved can remain a bystander. Businesses, developers, and regulators alike must engage in conversations that can pave the way for a future where AI can be leveraged safely, ethically, and commercially.
____________
Written by: Ateendriya
