What AI deepfakes mean for the future of commercial insurance

Takeaways:

• AI deepfakes lead personalities to trademark themselves for legal protection
• Development of coverage will parallel cyber breach insurance
• Insurers will need more details to price deepfake risks

The AI-powered acceleration of deepfake videos — convincing fabrications of real people —is reshaping how insurers handle reputational risks to celebrities and corporate executives.

In January, Matthew McConaughey trademarked several videos and audio clips of himself, including saying "alright, alright, alright." In April, Taylor Swift trademarked her image and samples of her voice. 

Gittings Photography

Trademarks are a new legal strategy to address deepfakes, but the insurance industry could start offering more specific protections beyond standard cyber and commercial specialty insurances, according to Courtney Horrigan, partner in the insurance recovery group at the Reed Smith law firm.

Coverage for damages suffered due to AI-generated deepfakes will likely develop similarly to how cyberbreach insurance began as an endorsement added to errors and omissions (E&O) insurance, but later became a dedicated product, Horrigan said. After insurers began writing policies to cover cyber-related losses, overall commercial insurance no longer was sufficient.

If insurers do the same with deepfake-related losses, Horrigan said, "They will stop being silent coverages. You will need explicit coverage under a policy for an AI risk."

For such coverage to become widely available, insurers must be able to quantify the risks of reputational damage and the resulting losses from that damage. This can fluctuate extremely in arts and entertainment, as Corrigan explained.

"The difficulty that the insurance industry is having as a whole in quantifying the risks of AI and valuing the artistic content of a creator, is an issue that you see oftentimes with art insurance, where third-party appraisers are trying to predict what the value might be of a certain creator's content," she said. "Of course, what we see with social media is things can go viral and the value of somebody's content can change very quickly."

If deepfake insurance evolves similarly to cyber insurance, it will face issues quantifying risks and value of losses, just as cyber insurance did. 

For instance, cryptojacking — breaching a cybersecurity defense to use the target's computing resources to mine cryptocurrency — had to be quantified as a risk and have losses more clearly defined. In 2017, cryptojacking became prominent, and it took one or two years for insurers to begin explicitly covering it in policies, according to Harrigan.

"What I see with respect to the potential for deepfake liability insurance coverage is the same type of evolution," she said. "You will see claims. You will see people arguing on both sides as to whether it's covered under the language and what the scope is. As the insurance companies see more claims, then they will be able to better underwrite for those types of liabilities. You'll get more clarity wrapped around the coverage."