Data security, privacy and copyright

- Published: January 9th 2024. Updated: January 12th 2024.
Data security
This article has been translated from Norwegian using artificial intelligence.
Data security is a collective term for methods, services, and tools that ensure that digital information and digital systems do not become unavailable, lost, read, or altered without the owners' consent or prior action (SNL, 24.10.23)
When using artificial intelligence, you must be very aware of what type of information you give the tool access to. If you do not take sufficient precautions, the data you enter into the writing robots may, in the worst case, end up in the wrong hands and be misused.
Personal data
Personal data is all information and assessments that can be directly or indirectly linked to you as an individual. Personal data can identify individuals alone or in combination with other data, such as name, address, date of birth, social security number, email address, and telephone number. Some information, such as health information, requires extra caution and is an example of special categories of personal data (formerly called sensitive personal data).
In Norway, the General Data Protection Regulation or GDPR (General Data Protection Regulation), which is a law adopted by the EU, applies. The General Data Protection Regulation regulates how personal data can be stored and used. GDPR provides a common standard for data protection in the EU and EEA countries, it gives individuals a number of rights and requires companies that process personal data to comply with strict data protection requirements.
The writing robots contain large amounts of data and information. When you enter text into the open writing robots, the text is stored permanently and used to train the robot further. This means that you cannot delete text or images you have entered.
Privacy is about protecting other people's personal information, and it means that you must not enter personal information into open writing robots. As a student, VID recommends that you use GPT UiO as a privacy-safe writing robot. This solution safeguards privacy with features such as Feide login. You can also be sure that your data will not be used to train the models further, and the name of the person using the service will not be disclosed. In addition to privacy, your data security is also taken care of.
For more information, see: Guidelines for the processing of personal data in research and student assignments at VID scientific university college guidelines
Copyright
In addition to ensuring that information does not go astray, you must protect the copyright of the material you give the writing robot access to.
If you have access to a scientific article that is in a database that requires login, you cannot post it on the internet or in a writing robot. The article is protected by copyright and must not be shared without the copyright holder's consent. If you give the writing robot access to the article, others may gain access to it by the article being built into the tool and becoming available to all users. If you violate copyright, this may have both legal and financial consequences.
It is also important to remember that writing robots are trained on large amounts of data. In many cases, it is unknown what data this is, and there may be problems related to whether the use is legal. Several foreign media outlets have blocked ChatGPT because the owners did not want their content to be used to train the writing robot.
Be careful
- You must have sufficient rights to the information you use. Source protection is important, and you must not share sensitive material or personal information.
- Always evaluate the text before entering it into a writing robot, cf. data security, personal data and copyright.
- We can use writing robots as tools, but the text must always be quality assured and processed through ordinary source work.
- Always critically evaluate the answers you get from a writing robot as they can give misleading and incorrect answers, contain built-in biases and stereotypes that are not easily apparent.