In a perfect world, organizations could circulate emails containing confidential information – financial data, new product developments, or other intellectual property – to stakeholders without risk of it ending up in the hands of competitors or the evening news.
Unfortunately, that’s not the modern world we live in. Insider leaks are commonplace, and discovering the source of a leak is notoriously difficult. Just look at last year’s Supreme Court leak; eight months and 126 interviews later and the leaker still hasn’t been identified. That’s not surprising. Most leakers elude capture because they believe – correctly – that they won’t be identified.
With little to deter leakers from distributing confidential content, leaders respond by sharing less.
For an organization, fear of leaks creates a chilling effect that results in watered down communications lacking in transparency and depth. And without the information they need to do their best, employees aren’t able to do their jobs as effectively.
What if the source of an organizational leak could be quickly and easily identified? That’s the promise of AI rephrasing, a new approach to confidential data protection.
Powered by generative AI and Large Language Models (LLMs), AI rephrasing allows the sender to easily construct uniquely phrased content for each recipient. Let’s take email, for example. Whether you’re sending 10 emails or 150,000, AI rephrasing personalizes each version while remaining semantically the same.
In an email, you might write:
"Thanks for the great discussion today."
"I want to thank each of you for a great discussion today."
"Thank you all for the great discussion today."
In this example, the alternate phrasing options retain the original meaning, though each is unique.
AI rephrasing is an effective deterrent and prevention tool because it can pinpoint precisely the source of an insider leak. In some cases, the leak may not be nefarious – perhaps information is carelessly handled or a computer is compromised. Either way, AI rephrasing gives organizations the ability to identify the source and remediate it quickly.
The idea of distributing slightly altered emails to a large group in hopes of revealing a leak isn’t new. In fact, Tesla founder Elon Musk famously did this in 2008 using “emails coded with one or two spaces between sentences” to try and identify an employee who leaked the confidential Tesla data and sold it to a news outlet.
What’s changed is the sophistication, speed and potential for automation that's now available with generative AI and LLMs. Within the body of an email, for example, the number of permutations can reach the level of astrophysics — so it doesn’t matter how few or how many emails are being sent – and it can do without any additional effort or time on the part of the sender. Plus, the technology is such that it survives end user reformatting or retyping into a personal computer.
The technology also recognizes that not all content is a candidate for rephrasing. There are certainly scenarios when text must not be altered. Senders have the option to approve the various phrasing options or to “freeze” certain lines of text – ideal in cases of quoted content that should remain unaltered (think: legal verbiage or a company’s precisely worded mission statement).
AI rephrasing, using our own purpose-built LLM is part of EchoMark’s comprehensive privacy protection suite that includes digital stamping and invisible watermarking. These tools allow organizations to share private information with confidence and create an environment where everyone is empowered to be better information stewards without impeding access to the flow of vital information.
Organizational leaders are often deeply worried about keeping confidential information private, and this fear reduces what they would otherwise want to communicate with employees and stakeholders. When trust breaks down and communication is restricted, individuals can’t do their best work and the organization becomes far less efficient.
Rather than breaking down trust, the reality is that technologies like AI rephrasing actually create an increased sense of stewardship among recipients. Care and confidence rise, and more communication can take place between people worthy of trust so that everyone involved can do their best work.
