Big Tech Scraps AI Usage Rankings Amid 'Token-Maxxing' Backlash

*Editor's Note: 'AI PRISM' (Personalized Report & Insight Summarizing Media) is an AI-based personalized news recommendation and summary service developed with support from the Korea Press Foundation. It curates and provides six tailored news articles for each reader type.

1. "We Just Wasted Tokens": Big Tech Scraps Usage Rankings

- Key Summary: Amazon abolished its 'Kiro Rank' system, which had evaluated employees based on usage of its AI code-generation tool 'Kiro.' Computing costs surged as employees instructed AI to perform unnecessary tasks to boost their evaluation scores. Dave Treadwell, Amazon's Senior Vice President of Engineering, said, "It was a leaderboard created with good intentions, but inflated token usage drove up cost burdens." Meta also discontinued 'Claudenomics,' a leaderboard that tracked token usage by 85,000 employees, while Uber burned through its annual AI budget in just four months due to token costs poured into AI coding services. In response, Salesforce introduced the concept of Agent Work Unit (AWU), which evaluates the volume of work processed by AI agents (autonomous AI assistants) instead of token counts, leading the industry's reorganization of performance measurement standards.

2. Why Airbnb, the Home-Sharing Pioneer, Turned to Hotels After 18 Years

- Key Summary: Airbnb officially launched a hotel booking service at its San Francisco headquarters on May 20 (local time), declaring its largest business transformation since its founding 18 years ago. Airbnb CEO Brian Chesky said, "There are times when you try to find a place to stay through Airbnb in New York and there's nowhere available," citing the shortage of home-sharing supply as the backdrop. The service launched in 20 major cities worldwide including New York, Paris, London, and Madrid, excluding large chain hotels and curating only boutique and independent hotels. The company also introduced a best-price guarantee that compensates the difference in credits if a booking is found cheaper on other reservation sites, along with up to 15% credit rewards on payments. Meanwhile, Uber also partnered with Expedia in the same month to add booking functions for more than 700,000 lodging facilities within its app, intensifying platform expansion competition.

3. "Level 3 at Less Than Half Tesla's Price": BYD Kicks Off Autonomous Driving Mass Adoption

- Key Summary: BYD Chairman Wang Chuanfu announced at a press conference in Shenzhen, China on May 28 (local time) that the company would supply its in-house developed Advanced Driver Assistance System (ADAS) 'God's Eye' for 12,000 yuan (approximately 2.7 million won). Vehicles equipped with this system will be priced under 100,000 yuan (approximately 22 million won), significantly lowering the price barrier compared to most Chinese ADAS-equipped vehicles currently priced above 150,000 yuan. BYD's 'God's Eye' is a Level 3 autonomous driving technology that allows drivers to take their hands off the steering wheel, one step ahead of Tesla's current Level 2+ capability and cheaper than Tesla ADAS's two-year subscription fee (approximately 3.57 million won). BYD also declared it would bear all costs in the event of an accident occurring while the technology is operating, clearly signaling its determination to popularize autonomous driving.

4. MS Vice President: "Mythos Is an Opportunity, Not a Threat… Catching 'Shadow AI' to Strengthen Security"

- Key Summary: Marco Casalaina, Microsoft's Vice President of Products and Core AI, said in an interview with Seoul Economic Daily on May 27 (local time) that Anthropic's AI model 'Mythos' is "like a double-edged sword" but "will be an opportunity, not a threat." Mythos surpasses Anthropic's existing top model 'Claude Opus 4.7' and is regarded as the most powerful ever in detecting cybersecurity vulnerabilities buried for decades. Considering its potential impact, Anthropic has limited the distribution of preview (trial) models to major big tech and financial firms, and Microsoft is collaborating to use Mythos to detect potential threats and block security vulnerabilities. Casalaina also emphasized the use of MS security tool 'Defender' to block the 'Shadow AI' phenomenon, in which employees personally use external AI models without official company approval.