Major tech and AI companies are set to introduce watermarking systems for AI-generated content, following a pledge made to the White House.
The move aims to address concerns over the spread of misinformation and biased content produced by AI tools.
Seven leading tech and AI companies, including Google, Microsoft, Meta, Amazon, and generative AI firms OpenAI, Anthropic, and Inflection, have agreed to implement "voluntary commitments" to improve the safety and accountability of their products.
These commitments include developing systems to identify and watermark content generated by AI tools, to help distinguish between human and AI-generated content.
The use of AI to produce convincing text and images has raised concerns over its potential to spread misinformation and perpetuate biases.
In May, a fake image of the Pentagon circulated on social media, causing markets to briefly drop, while a recent political ad in support of Florida Governor Ron DeSantis used AI to replicate Donald Trump
's voice for a soundbite that never actually occurred.
Research has shown that AI-generated content can be difficult to distinguish from human-written content, with a study finding that a majority of people were unable to tell the difference between tweets written by humans and those generated by ChatGPT.
As a result, the development of watermarking systems for AI-generated content has become a priority for many tech and AI companies.
Google, for example, has already integrated watermarking and is developing its "About this image" tool, which will allow users to identify where an image originated and whether it has been featured on fact-checking sites or news publications.
Meta and Inflection, two other companies involved in the White House pledge, have also committed to developing similar systems.
While the use of AI in content creation has the potential to produce powerful and convincing content, the industry is still in its early stages of developing safe and trustworthy AI.
Inflection, the studio behind the personal AI chatbot "Pi," has acknowledged that there is still much work to be done, describing the agreement with the White House as "simply a springboard and catalyst for doing more." In conclusion, the introduction of watermarking systems for AI-generated content is a significant step towards address