As generative AI tools become more sophisticated and easier to use, the line between inspiration and infringement is becoming increasingly hard to discern. Is AI a power for good when it comes to improving our lives and productivity, or is it a malign force that bad actors can use to imitate, steal, defraud, and much more besides?
Online brand protection
Before diving in, it would make sense to look a little deeper into the issues of intellectual property (IP) and brand safety online. Did you know that illegal streaming alone costs the U.S. economy almost $30 billion every single year? However, this is far from the only issue. Written content, images, logos, specific design elements, and other brand assets are the main elements of a company’s identity and a source of significant risks that can directly affect business revenue in case these assets are abused.
If a business fails to protect its brand assets and IP, it has to contend with at least two major issues: financial losses (imagine a loss in revenue due to the proliferation of product “dupes”) and disrupted SEO results. For instance, duplicated content, which may account for up to 60% of the internet, is usually treated negatively by search engines, reducing brands’ visibility. Which, in turn, might also result in financial losses.
Businesses that aim to fight thieves and fraudsters have started to collect web intelligence at scale as their go-to defense against counterfeits and unauthorised use of brand assets. Web scraping tools, such as scraper APIs, have been used for years to systematically comb through the internet to spot instances where protected content is being used without permission. For example, with dedicated scraping infrastructure, companies can gather localized, real-time product data from various e-commerce marketplaces that can be used to sell stolen or counterfeited products.
That’s not to say, however, that web scraping is free of deficiencies. Proxies can be blocked, the whole infrastructure requires constant human oversight, and smaller companies struggle to keep up with the sheer volume of dynamic online content. This is where AI comes in. AI and machine learning have completely changed the game when it comes to IP protection. Far from being just a minor upgrade, it’s a whole new way of thinking about how to best protect the business assets and gather web intelligence for defence.
The frontiers of AI copyright protection
Advancements in deep learning have given image recognition algorithms a massive boost in terms of accuracy. They can now more confidently scan e-commerce platforms, websites, and social media sites to detect counterfeit logos, packaging, and products. Systems like IBM Watson and Adobe Sense are sophisticated enough to spot unauthorised usage, even if the images have been slightly tweaked, which is a big improvement over traditional pixel-matching techniques.
A typical use case would be deploying an AI system to keep a constant watch on marketplaces online for counterfeit goods or logos. Upon detection, the system immediately springs into action, flagging the listings that feature the offending items. Superior to the usual reactive methods, this forward-looking approach is effective not only at preventing loss of revenue but also at putting up barriers to brand dilution.
While the recent trend of focusing more on visual content is largely justified, text remains important, too. Thanks to cutting-edge NLP models like Amazon Comprehend, linguistic analysis has never been more accurate. These algorithms can analyse everything from product descriptions to reviews and website content, accurately identifying unauthorised reproductions. Most impressive of all, modern NLP systems are capable of detecting not just direct copying but also paraphrased content, often used to dodge traditional plagiarism detection.
AI has also enabled developers to create unique digital fingerprints for audio, video, and text content. Acting as unique identifiers, they can automatically detect unauthorised use across the internet.
Finally, AI can be beneficial when analysing infringement patterns to predict future violations. For instance, if analysis shows that a certain region or platform is rife with copyright infringement cases for a specific brand, businesses gain the latitude to implement preventative measures tailored to that region or platform.
However, businesses shouldn’t be fooled into thinking that AI tools will suddenly solve all their brand protection issues without any additional effort. Despite quick advancements, AI systems still have precision issues and lack contextual understanding. For example, by indiscriminately flagging brand-related content on social media networks, such as Instagram or Facebook, AI systems harm bloggers or small UGC influencers who are mentioning certain brands only for informational purposes or make genuine product reviews for their target audience.
Two sides of the same coin
To sum up what was already said, in the right hands, AI can be a powerful ally in protecting intellectual property and brand assets. Unfortunately, it can also have a negative impact on creators or even be used by malicious actors.
Large Language Models (LLMs) trained on vast amounts of human-generated content are already quite adept at churning out passable articles, stories, and copy. The problem is, this seemingly original content often includes some of the data these models were themselves trained on. And this, in turn, raises some fundamental questions regarding authorship: If an algorithm trained on copyrighted works creates new works featuring the same themes and style, does that constitute infringement?
We’re only just starting to see courts and lawmakers really grapple with the ways in which machine-generated content fits into copyright law. Some big-profile cases include The New York Times’ suit against Microsoft and OpenAI in 2023, and Sarah Silverman against Meta in the same year. Thus, for visual artists, photographers, and designers, AI still represents a big challenge — not only in terms of their livelihoods, but also of creative control.
The issue haunts not only artists, though. Some marketers, SEO specialists, and digital content creators employ AI tools to produce low-quality content on a large scale, often using content materials created by other brands. Further, with the help of AI, it is also easier to create social media accounts and entire websites for selling counterfeit goods or even fraudulent shops offering items that simply don’t exist.
The truth is that AI is neither doom and gloom nor a panacea to all the world’s ills. In the end, it will be human intervention that decides the question. To strike the right balance, we’ll need not just clever tech fixes, but also updates to our legal frameworks, and a good dose of ethical thinking. Complicating a lot of this is the fact that copyright law was developed well before the digital era, and it’s largely incapable of handling the complexities of today’s data market and AI technologies. Can this be fixed?
The first steps could include broadening the currently existing fair use guidelines by including specific AI training activities. At the same time, legislators should reinforce the need for proper attribution and come up with new licensing models tailored for AI systems. It also bears pointing out that most of these guidelines and regulations will amount to little if they’re not extended internationally, as AI cares little for borders between nations. Making good on these worthy aims will likely include collaboration between AI developers, content creators, and legal experts.
In summary
For businesses aiming to secure their brand assets, AI has much to offer in terms of wide-scale online monitoring, preventive flagging, and even striking back at fraudsters. Recent advancements in image recognition, NLP, and pattern detection equip legitimate IP owners with the means to protect their creative works and brand identities.
However, to make sure that AI strengthens rather than undermines the position of IP owners, which is key for both creative and commercial businesses, serious change will have to happen. This includes developing and implementing well-thought-out technical safeguards, smart regulatory adjustments, and extensive collaboration between stakeholders.
