Meta Imposes Invisible Watermarks on AI-Generated Images to Curb Misuse
Summary:
Meta, the company previously known as Facebook, plans to add invisible watermarks to all images created with its artificial intelligence (AI) tool to prevent misuse. Unlike traditional watermarks, these will resist common image manipulations and cannot be removed easily. Initially applied to images produced by Meta AI, the watermark service will later extend to other Meta services using AI for image generation. The 'reimagine' feature for Facebook Messenger and Instagram, which permits users to exchange AI-driven images, will also receive this feature. This change responds to the surge in AI-aided scams that fabricate misleading content, as seen with fake videos, audios, and images that have circulated the web.
Meta, the company formerly known as Facebook, has announced plans to embed hidden watermarks in all images created through artificial intelligence (AI), as part of their strategy to thwart the misuse of the tech. According to an update provided on December 6 regarding Meta AI - the firm's intelligent assistant - an invisible watermark will soon be incorporated in every AI-generated image stemming from the 'imagine with Meta AI experience'. Just like other AI chatbots, Meta AI fabricates images and content based on users' input; yet unlike its counterparts, Meta intends to discourage malfeasance by refusing to let its resource be exploited as a tool for deception.
The incoming invisible watermark feature will be resilient to attempts for removal by creators. “In the forthcoming weeks, a hidden watermark will be integrated with the 'imagine with Meta AI experience' for enhanced transparency and accountability," Meta has stated. A deep-learning model will be used to affix watermarks - undetectable by the human eye - to pictures fabricated via the AI tool. However, the concealed watermarks can be identified with an appropriate model. Conventional watermarks are sub-par to Meta's proclaimed AI watermarks, known as 'imagine with Meta AI', which are robust against usual image modifications including cropping, color alterations (brightness or contrast adjustments), screenshots, and more.
Initially, the watermark service will be applied to pictures produced through Meta AI, but there are plans to extend the feature to other services that employ AI in image generation. In addition to this, Meta AI has launched the 'reimagine' feature for Facebook Messenger and Instagram. This update facilitates the exchange of AI-driven images among users and will integrate the hidden watermark feature within both messaging platforms.
AI platforms, such as Dall-E and Midjourney, already offer the option to incorporate traditional watermarks in their content. However, these can be easily removed by cropping the image peripherally. Certain AI tools can even eradicate watermarks from images automatically, a feat that Meta AI asserts will be impractical with its product.
Following the popularization of generative AI tools, a variety of business bodies and celebrities have denounced AI-aided scam ventures. Scammers exploit handy tools to produce falsified videos, audios, and images of renowned people, which they then proliferate across the web.
In May, an AI-concocted picture depicting an explosion close to the Pentagon, headquarters of the US Department of Defense, momentarily led to a dip in the stock market. This fabricated image was subsequently shared and spread by other news publications, thereby causing a cascading effect. Nevertheless, local law enforcement, including the Pentagon Force Protection Agency, accountable for the building's security, refuted the circulating reports, emphasising there had been "no explosion or incident."
That same month, Amnesty International, a human rights organization, was duped by an AI-generated image purporting to show police oppression, which they subsequently used for campaigns against law enforcement. Erika Guevara Rosas, Americas director for Amnesty, acknowledged the error, saying "We have retracted the images from social media updates, since we don’t want the dispute over the use of AI-concocted images to sideline support for the victims and their calls for justice in Colombia.”
Published At
12/7/2023 11:50:05 AM
Disclaimer: Algoine does not endorse any content or product on this page. Readers should conduct their own research before taking any actions related to the asset, company, or any information in this article and assume full responsibility for their decisions. This article should not be considered as investment advice. Our news is prepared with AI support.
Do you suspect this content may be misleading, incomplete, or inappropriate in any way, requiring modification or removal?
We appreciate your report.