AI promises to revolutionise much of the IP space, yet concerns around the security and trustworthiness of AI and the barriers to entry bring challenges for trade secret management, writes Seurat IP Manager Laura Daly
AI promises to revolutionise much of the intellectual property space and make corporate IP departments more efficient. However, there is a lot of hesitancy to embrace the new technology, particularly for small organisations who are still seeing high barriers to entry.
LLMs and other AI models show much promise for helping companies secure their trade secrets. The IT department could potentially use AI tools to monitor and detect patterns and behaviours of access that may cause risk, for example detecting unauthorised intrusions or anomalous access, enabling pre-emptive action before a breach occurs. Right now, this type of analysis is prone to human error, and due to capacity issues is often limited to a post-breach failure analysis.
Tools are also increasingly being marketed to IP departments. Tools that promise to streamline internal processes, for example by automatically tagging trade secrets as sensitive information and sorting them into appropriate ‘buckets’ depending on what is being protected. Other tools are also being developed to monitor breaches and detect misuse via publicly available information.
So why the hesitancy to adopt? The general concerns around the security and trustworthiness of AI and the barriers to entry to, particularly small businesses, make the spin-up challenging. Many of the recent entrants into this space have built platforms on the Open AI Chat-GPT platform. Open AI states that “[n}o customer data or metadata [is used in the] in training pipeline for…ChatGPT Enterprise customers”, however, it is difficult to ensure that start-up AI IP management and monitoring companies have the correct agreements and licences in place to protect their customers’ interests. Without being convinced of the privacy of their data, many companies are hesitant to proceed.
Larger enterprises may be able to spin-up their own tools build on the Open AI platform with appropriate licensing and protections, but smaller and start-up businesses that barely have a legal team never mind a legal engineering team simply lack the resources.
Companies (even start-ups themselves) may be hesitant to spend their IP budget on tools from new entrants to the field that may well run out of runway or fail to keep promises as to functionality and security. As the more successful companies are bought out by the larger players in the space, as has previously happened with IP landscaping, IP management and contract management start-ups it may be that legal management has more confidence to invest in AI trade secret management tools both from a compliance perspective and a return-on-investment perspective.
Particularly in the legal space there are real concerns about the effectiveness and accuracy of AI. There have been several high-profile cases of attorneys being sanctioned over false information derived from AI included in filings. Management of trade secrets by AI depends on the accuracy of algorithms and data input. Errors or biases in AI can lead to improper handling of sensitive information.
Whether a company makes an AI tool for handling trade secret information in-house, or spins up a commercially available tool, those biases, inaccuracies, and the possibility of ‘hallucinations’ remain a concern. While many professionals in the IP space have a technical background, few have the level of expertise required to test and verify that the output from an AI is accurate and the method leveraged by an AI system to produce the output is accurate and reproducible.
Every day there appears to be an increasingly bewildering number of new options for implementing AI tools to assist legal departments. Even down-selecting to an appropriate tool or a decision to make versus build requires a not insignificant use of bandwidth by teams that could be more usefully applying their skills elsewhere. In-house IP departments are often highly constrained in terms of resource. Spinning up an internal project or implementing an external tool can require significant resources both to ensure that data input is accurate and as clean as possible and to make the implementation a seamless process. Many IP departments have experiences of tools where the implementation was done so badly as to render them practically unusable.
There are also cost implications, IT implications – both from an implementation perspective and a security perspective – and regulatory concerns. Done well, AI based trade secret management has much promise to leverage new technology to assist IP professionals in their day-to-day job. However, it is critical that it is done well. The damage caused both on an operational and reputational basis by poor implementation could cause significant setbacks.
IP professionals by the very nature of their training tend to be a risk-averse bunch. AI tools for monitoring trade secret compliance and taking on other legal tasks show much promise to increase the efficiency of IP teams. Barriers to entry are difficult to overcome, however, particularly for smaller IP teams.
If a company chooses to build their own platform on Chat-GPT or another LLM there is an associated cost, as well as a requirement for engineering resources that may not be available. The risks around data privacy and compliance are the same, if not worse, than using a commercial solution.
If a company leverages outside commercial solutions, there are a plethora to choose from, which makes it hard to sort the valuable wood from the trees. Companies are often reluctant to spend money and time bringing in tools from start-ups that may not be around in a year, may not fulfil their promises and may have even worse compliance and privacy issues when there is no ready access to the black box they have built their tools on.
In ten years, once names and reputations have been established and the consequences and limitations of AI are better understood, IP portfolios and their management will have evolved beyond recognition. However, right now many of us are dipping our toes in the pool, but we are concerned by the sharks we cannot see in the water.