AI Generated Text – Security and Copyright
Kindus has already discussed the nature of AI generated text how bad grammar is no longer a flag that emails and documents are to be not to be trusted.
Now consider the threat posed to Intellectual Property and privacy from AI generated text. With the basic model of ChatGPT established other similar engines have been released, some free some running on subscription models. It is unlikely that the genie is going to let itself be put back in the bottle.
The AI engines produce a block of text but omit any provenance. An academic article will include a list of references to at least the background reading behind that article. Many web pages including this blog contain links to other pages indicating what the text is based on. The very nature of the web is that users should be able to follow through links, find out more and decide for themselves what can be trusted. In pieces such as this much of the text is the opinion of the writer. There are few external links because it has been created from scratch by a human. The majority of web pages are free to access (subscription newspapers being a notable exception) and other sites may or may not link into Kindus’ articles without restriction. Many sites work with advert hosting engines which will pay for page views or click-throughs encouraging the site owner to establish back links. Others including Kindus do not. We do, however, hope that the reader enjoys these short blog posts and might take some time to explore what Kindus does and the services we offer.
An old fashioned Google search will take the user to a site that they might be interested in and from the URL and details on the page they will have some idea who created its content. That whole process is sidestepped by the AI text solution. Anyone earning their living or hoping to derive some good from writing content will not receive any credit from AI creations. Taking this model to its unlikely extreme: The creation engines rely on content that is already available and created by others. If there is no incentive to create new content the engines will have less data to rely on and their content in turn will lower in quality. While this scenario is improbable the effect of AI text on creativity can only be negative.
There are security as well as copyright concerns as there are no clear rules as to what the text generator might unearth. While it is an obvious rule that nothing should be put onto an open Internet site that should be kept secret it is wise to assume that mistakes might be made. Fraudsters will often find useful information by combining data from several seemingly harmless and unconnected sources. For example corporate emails can be guessed from names even loosely associated with information from that organisation together with likely corporate email formats. AI generated text might be used with sophisticated search criteria that can expose information by combining results from many sites into a plausible whole.
As a final step Kindus put this article title ‘AI Generated Text – Security and Copyright‘ into the AI blog creator Vondy. The generated text covered some of the ground from our AI text article and discussed to what extent copying data is ‘fair use’. The concept of the rights and privileges of the original owner was not brought up.
