AI This Week

Meta has frozen hiring for its artificial intelligence teams as it reassesses resource distribution. The pause affects roles in generative AI and the team behind Llama, Meta’s open-source large language model. This decision comes despite CEO Mark Zuckerberg’s emphasis on AI as a top priority. The company continues investing in AI infrastructure, including plans to acquire 350,000 Nvidia H100 GPUs by year’s end.

A new analysis by Appfigures reveals that ChatGPT’s mobile app has accumulated $2 billion in global consumer spending since its launch in May 2023. This figure dwarfs the combined lifetime revenue of competitors like Claude, Copilot, and Grok by 30 times. In 2025 alone, the app has earned $1.35 billion, showing a staggering 673% year-over-year growth. The app generates $193 million monthly on average, with a lifetime spending per download of $2.91—significantly higher than its rivals. Additionally, ChatGPT leads in installations, surpassing 690 million downloads globally. The U.S. market accounts for 38% of its revenue, while India leads in installations. Despite emerging competition, ChatGPT continues to dominate the mobile AI landscape.

The U.S. electrical grid is straining under the explosive demand for colossal AI-driven data centers. According to a report by JLL, data center vacancy rates have hit a record low of 2.3%, as facilities now require power on scales equivalent to entire cities. Hyperscale sites are demanding gigawatts of electricity and thousands of acres, overwhelming utility systems with speculative applications. Northern Virginia and Dallas account for 50% of new demand, while developers scramble to reserve capacity for unbuilt facilities. Predictions point to $1 trillion in data center developments in North America from 2025-2030.

Anthropic has introduced a new feature in Claude Opus 4 and 4.1, enabling the AI to end conversations in rare, extreme cases of harmful or abusive interactions. This decision stems from exploratory work on AI welfare and broader efforts to improve model alignment and user safeguards. During testing, Claude exhibited a strong aversion to harmful tasks and signs of distress when faced with requests tied to violence, exploitation, or abuse. The feature activates only as a last resort after multiple redirections fail or when explicitly requested by a user. While rare, these interventions highlight a commitment to mitigating risks without compromising user experience. Conversations can still be restarted or revisited through edits.

MIT researchers have harnessed generative AI to develop groundbreaking antibiotics targeting drug-resistant bacteria. The team successfully designed compounds effective against Neisseria gonorrhoeae and MRSA, pathogens linked to millions of deaths annually. Using AI, they screened over 36 million theoretical molecules, discovering structurally unique antibiotics that disrupt bacterial membranes. One standout candidate, NG1, proved potent in fighting gonorrhea infections in mouse models. A second study used an unrestricted AI approach to design 29 million molecules, pinpointing DN1 as a promising compound against MRSA. Both antibiotics are now undergoing refinement for further testing. This research opens the door to exploring untapped chemical spaces and combating antimicrobial resistance with fresh strategies.