Could Watermarks Help Users Navigate AI-Related Threats?

Many individuals are concerned about the future of AI, including the White House, which encouraged companies like Amazon, Anthropic, Google, Inflection, Meta, Microsoft, and OpenAI to commit to helping with the management of artificial intelligence. Other companies, including Adobe, Cohere, IBM, Nvidia, Palantir, Salesforce, Scale AI, and Stability, have joined in this pledge to maintain “the development of safe, secure, and trustworthy AI,” according to the White House.

Why is this commitment such a big deal? Let’s explore this idea in today’s blog.

Imagine AI-Generated Content with Watermarks

Artificial intelligence is remarkably interesting and helpful in certain contexts, but it’s also a tool that cybercriminals can use against unsuspecting victims. Tools can be used to create deepfake images and replicate voices to scam victims, not to mention the plethora of other dangerous ways it can be used against innocents.

The current administration is seeking to push these companies to create a technology to watermark AI-generated content, placing a label on the content so viewers can determine what platform was used to create it. In theory, the watermark should allow users to identify content created with AI, further assisting them in identifying potential threats and scams.

Furthermore, there are other safeguards on the table, including the following:

• Tech companies will invest in cybersecurity to protect the data that powers AI models.
• Independent experts will be responsible for testing AI models prior to their public release.
• Companies will research the risks associated, and how they could impact society at a large scale, including how bias and inappropriate use could factor in, and flag behavior deemed problematic.
• Third parties will have an easier time discovering vulnerabilities and report them when they are addressed.
• These companies will share risk-associated data with others, including society and academic researchers.
• These firms will disclose security risks, including those of their own products, to society, along with their biases.
• These firms will develop AI that can handle some of the world’s more challenging issues.

All of this said, there are no standards or practices that are enforceable by the government in this realm, but an agreement—even a potentially empty one—could be enough to get the ball rolling on certain AI-related issues.

Let Us Help Your Business

We dedicate ourselves to helping our clients navigate the confusing and perilous world of cybersecurity, and technology in general. To learn more about what we can do for your business, call us today at (02) 8825-5555.