Vercel’s analysis underscores a critical gap in AI endpoint security: the high cost of inference makes it a prime target for exploitation. While per-request verification, as Vercel suggests, may mitigate some risks, it’s unclear whether this approach can handle the scale and sophistication of evolving attacks. The reliance on residential proxies and adapter tools highlights how attackers are outpacing traditional defenses.
As AI becomes more integrated into workflows, the industry may need to rethink security architectures entirely, not just layer on new checks. The challenge lies in balancing robust protection with the seamless user experience that AI promises.