Why Was ChatGPT Losing Its Users - The OpenAI Scandal Explained

The image used in the cover of the article has been taken from: https://www.linkedin.com/pulse/visionaries-ai-sam-altman-dario-amodei-future-artificial-intelligence-rptcf.

ChatGPT is one of the most widely used AI platforms with around 900 million users. Recently reports of boycotting ChatGPT with websites like QuitGPT running these movements have indicated loss of around 2.5 million users. Reports suggest that users have cancelled subscriptions and have switched to Anthropic’s Claude. What has triggered this level of response?

What is Pentagon asking for from its AI contractors?

Before OpenAI, Anthropic was the US Department of Defence's contractor. After a meeting between Dario Amodei (CEO of Anthropic) and Pete Hegseth (US Secretary of Defense), it was surfaced that Hegseth asked Anthropic to accept "any lawful use of its tools". Should Anthropic refuse Pentagon's request, Hegseth threatened to remove Anthropic from the department's supply chain and to label it a "supply chain risk". This label affects the company's consideration for other government contracts. While majority of Anthropic's revenue is not from government contracts, the threat still risked the loss of billions of dollars and reputational harm for the company. As of 4 March, Anthropic was the first US company to have been officially labelled as a "supply chain risk" — its executives have moved a lawsuit to prevent Pentagon from putting it on a security blacklist.

On the same day that Anthropic was deemed a security risk, the department also announced a contract with OpenAI that seemed to replace Anthropic's model, Claude. This led to backlash from its employees and outcry from the public, as it seemed OpenAI was willing to cross red lines that Amodei refused. Caitlin Kalinowski (who was a member of OpenAI’s technical staff overseeing hardware) left the company. She cited that she was leaving not because of the people at OpenAI, sharing that she had great respect for Sam Altman and the team, but objected to the principles. It seemed that the deal was rushed, and made OpenAI look “opportunistic and sloppy”.

What does this mean for the AI platform's users?

The democratisation of AI has brought about speed and convenience in every aspect of our lives, including the ability to conduct surveillance. A 2019 American Civil Liberties Union report highlights how video analytics powered by AI allows machines to search for data within a video and to create real time analysis with it. In terms of surveillance capabilities, the AI could allow authorities to monitor individuals with unprecedented speed and accuracy, combing through mountains of digital data to track people’s movement and behavior. In addition, these platforms have been voluntarily provided access to a vast amount of user data. The concern is that a military contract between a company like OpenAI will enable Pentagon to access the capabilities of the AI and the data available without oversight in the name of 'national security'. These fears are not unfounded as reports have indicated that Pentagon demanded AI companies remove safety guardrails on their models to allow a broader range of military applications.

After the fall out - what happened next?

OpenAI was forced to renegotiate the agreement. Sam Altman, CEO of OpenAI, and Katrina Mulligan, head of national security partnerships for OpenAI used X to convey that OpenAI services will not be used by Department of Defense intelligence agencies (for example, the NSA). While these are claims being made, transparency in this regard is minimal — it is hard to ascertain to what extent the agreement allows OpenAI’s use in surveillance. OpenAI’s post states that as per the new agreement “the AI system shall not be intentionally used for domestic surveillance of U.S. persons and nationals”. What is the scope of ‘intentional use’? What are the exceptions when use will not be regarded as intentional and hence domestic surveillance will be allowed? These questions remain unanswered.

These deals signal towards the normalisation of AI deployment in ‘warfare and surveillance’, amongst other commercial or governance-related use cases. If there is something we can learn from this event, it is a resounding affirmation that corporations view user data as a currency to deal in. When we disclose our information to AI, be it uploading our pictures, videos, sensitive documents, you should ask yourself: Is it okay if someone gets access to this document? When you disclose personal details to AI, remember that for these systems it’s just data which, if can be used for you, then it can also be used against you.