Welcome to Daily Zaps, your regularly-scheduled dose of AI news ⚡️ 

Here’s what we got for ya today:

• AI bots now beat 100% of traffic-image CAPTCHAs

• AI agents from major SaaS companies

• AI can outperform human CEOs

• California’s safe and responsible AI efforts

Let’s get right into it!

Sign Up | Advertise | Tools

TECH

AI bots now beat 100% of traffic-image CAPTCHAs

New research shows that AI models, like the YOLO image-recognition model, can now achieve a 100% success rate in solving Google's image-based reCAPTCHA v2, which is still used on millions of websites. By training the model on 14,000 labeled traffic images and implementing additional steps like using a VPN and mimicking human-like mouse movements, the researchers created a bot that can consistently bypass the CAPTCHA system.

While previous studies achieved only around 70% success, this new research highlights that AI is now capable of surpassing human-level CAPTCHA-solving performance. As AI becomes more adept at tasks once thought to be human-exclusive, CAPTCHA systems will need to evolve, with a focus shifting towards more subtle identification methods like device fingerprinting.

BIG TECH

AI agents from major SaaS companies

Businesses are exploring how AI chatbots, particularly AI agents that can independently perform tasks like scheduling, can enhance operations, but skepticism remains. Tech giants like Salesforce, Microsoft, and ServiceNow have recently introduced AI agents that aim to automate tasks such as placing orders or scheduling meetings, but these tools still require detailed instructions and setup, limiting their current functionality.

Customers are cautious about trusting AI with significant tasks, as the technology is not yet advanced enough to operate without supervision. OpenAI, with its more capable agents on the horizon, is prompting these companies to accelerate adoption. However, a learning curve exists, and many customers are hesitant to fully integrate the technology into their workflows.

BUSINESS

AI can outperform human CEOs

While AI replacing a CEO might seem far-fetched, a recent experiment demonstrated that AI, specifically GPT-4o, can outperform human participants in strategic decision-making tasks. In a simulation of the U.S. automotive industry, GPT-4o consistently outperformed top human players in product design, cost control, and market responsiveness. However, it struggled with unpredictable events like market collapses, leading to its early dismissal by a virtual board.

This highlights AI's strengths in data-heavy, short-term optimization but also its limitations in handling long-term flexibility and black swan events. The experiment suggests that AI won’t replace human CEOs but can augment decision-making, enabling leaders to focus on strategy, ethics, and vision for sustainable growth.

GOVERNMENT

California’s safe and responsible AI efforts

Governor Gavin Newsom of California has recently signed 17 bills over the past 30 days addressing the regulation and deployment of Generative AI (GenAI) technology. These bills represent the most comprehensive legislative package in the country, tackling issues like deepfakes, AI watermarking, protecting children and workers, and combating AI-generated misinformation.

Some notable bills include SB 896, which requires risk assessments of GenAI threats to critical infrastructure, and AB 2655, which mandates the removal or labeling of deceptive AI-generated election content. Despite signing 17 bills, Governor Newsom vetoed SB 1047, citing concerns over its rigid standards that may hinder innovation.