Written by 2:05 pm Discussions

### Limitations in AI Company Promises for Safeguarding Consumers from Online Challenges

Indemnities offered by Amazon, Google, and Microsoft are narrow.

As part of a high-stakes bet on relational AI, major tech corporations are competing to offer innovative services such as artificial intelligence and virtual assistants.

The leading global cloud computing firms, known for introducing cutting-edge artificial intelligence solutions to their clientele, are currently providing only a limited set of protections against potential trademark disputes.

In a multibillion-dollar gamble on advanced AI systems capable of generating human-like text, images, and code rapidly, Amazon, Microsoft, and Google are in a race to introduce novel offerings like virtual assistants and chatbots.

These AI models are trained on internet-derived data like images and text, leading to concerns that rights holders, including internet companies and publishers, may take legal action against third parties utilizing AI tools trained on their copyrighted content.

While the top three cloud computing giants have committed to safeguarding consumers and businesses against intellectual property disputes, these legal safeguards primarily cover designs created by or under the supervision of Google, Amazon, and Microsoft, as outlined in the compensation clauses released by these companies.

Matthew Sag, a law professor at Emory University, praised these provisions as astute business strategies that instill confidence in users regarding permissible usage. However, Brenda Leong from Luminos Law emphasized the importance of understanding the precise definitions and limitations outlined in these provisions.

Microsoft, Amazon, and Google have refrained from making public statements on this matter.

Despite the availability of these tools on the cloud providers’ platforms, the protections offered to users do not extend to third-party models, such as those developed by AI startup Anthropic, which have received investments from Amazon and Google.

Only content generated by Amazon’s proprietary models, like Titan, and certain AI software from the company are covered by these provisions. Similarly, Microsoft offers protection for tools developed by OpenAI, with whom it has a significant partnership, and its internal models.

According to an intellectual property attorney involved in these issues, the inclusion of these clauses was prompted by a heightened awareness of legal risks among users.

The cloud service providers have been augmenting their tools with security filters to screen out potentially infringing content. While these measures aim to minimize risks, customers of Microsoft, Amazon, and Alphabet may seek more specific indemnities tailored to their requirements, a practice that is not yet widespread.

Legal battles related to relational AI, involving notable figures like Sarah Silverman and prominent writers, are ongoing, with companies like OpenAI and Meta facing allegations of improperly training employees using copyrighted material.

To address concerns over potential litigation, additional security layers have been introduced in contracts, acknowledging the prolonged resolution times for such disputes.

Google’s indemnification policy does not cover models fine-tuned by customers using internal company data, unlike Microsoft’s approach. This practice allows businesses to enhance basic models for more tailored results, with Amazon’s coverage extending to fine-tuned Titan models unless infringement stems from the customization.

Challenges may arise in lawsuits against users of conceptual AI tools rather than their creators, as evidenced by a case dismissed last year due to the difficulty in proving reliance on copyrighted training data for AI-generated content, as noted by US Judge William Orrick.

Visited 3 times, 1 visit(s) today
Last modified: January 10, 2024
Close Search Window
Close