OpenAI and New Policy: Focus on Privacy, Freedom, and Teen Safety

A recent tragic incident in which a young life was lost after a conversation with ChatGPT and following a parental lawsuit against OpenAI and Sam Altman has prompted the company’s owners to take certain steps.

A few days ago, they introduced a new feature, parental controls to ‘support families in setting healthy guidelines,’ and now they have announced new safety features focused on youth, specifically teenagers, on their corporate blog.

In his blog, Sam Altman explains the new principles and safety measures regarding the use of artificial intelligence, highlighting how three key goals sometimes conflict in practice: user privacy, freedom of use, and the protection of teenagers.

First, privacy

The company emphasizes that protecting privacy in conversations with AI (ChatGPT) is extremely important.

– People are increasingly using AI to discuss personal and sensitive topics, which makes this technology different from previous digital tools – Altman writes in the blog.

He further points out that conversations with AI can be as personal as those with a doctor or lawyer, and such information should have a special level of protection. Therefore, the company is working on advanced security features that ensure data privacy, even from OpenAI employees.

– However, certain exceptions exist, and automatic systems will monitor serious abuses, such as threats to life, plans to harm others, or planning mass cyberattacks, and such queries will be forwarded for human review – it is emphasized in the blog.

Second principle – freedom

The second key principle is freedom of use. OpenAI wants users to have as many opportunities to use the tools within safety boundaries. The models are being developed to become more flexible; for example, the default behavior of the model does not include flirting, but if an adult user requests such a conversation, the model can accommodate them. For more sensitive topics, such as writing fictional stories about suicide, the model will provide assistance in a creative context but not in the case of actual self-harm. Internally, this is referred to as ‘Treat adult users as adults,’ which means expanding freedom as much as possible without causing harm to others.

Third principle – protection

The third principle relates to the protection of teenagers. OpenAI places the safety of minors above privacy and freedom, as it believes that new AI systems require significant protection for vulnerable users.

– The company is developing a user age assessment system to distinguish individuals under 18 from adults (ChatGPT is intended for users aged 13 and older). If there is any doubt, a safer ‘under-18’ mode will be applied. In some cases, identification may be required, which is a compromise on privacy for adults but is considered justified for the purpose of protecting teenagers – the company states.

In translation, identification will soon be required for all users, although this is not currently indicated. However, when an identification measure is introduced for a specific purpose (e.g., protecting teenagers), there is a tendency for such requirements to expand over time to other cases or users. This does not mean it will happen immediately, but there is a strategic risk of the expansion of such requirements, which is a common topic in discussions about privacy and digital rights, similar to the European Chat Control.

Different rules for teenagers

Different rules will apply to teenagers. Flirting conversations or discussions about suicide and self-harm, even in the context of creative writing, will not be allowed. If a minor shows suicidal thoughts, the system will attempt to contact the parents, and in cases of immediate danger, the relevant authorities. OpenAI has also announced new parental controls that allow monitoring of children’s activities, blocking certain functions, and notifications about potential stress in teenagers.

The company acknowledges that the principles of privacy, freedom, and protection of teenagers are in conflict and that not everyone may agree with their decisions, but they emphasize that the decisions were made after consultations with experts and in the best interest of users.