Latest AI (Artificial Intelligence) News
OpenAI model reportedly solved an 80-year geometry problem
OpenAI’s internal model autonomously disproved a geometry conjecture that had stumped mathematicians for decades, with Fields medalist Tim Gowers calling it a milestone in AI mathematics. The report highlights growing evidence that frontier models are starting to contribute meaningfully to original scientific and mathematical discovery .
Anthropic co-founder predicts Nobel-level AI breakthrough within 12 months
Jack Clark said at Oxford that AI could help humans make a Nobel Prize-worthy discovery within a year. He also forecast bipedal robots assisting tradespeople within two years and warned of a significant risk from recursive self-improvement by 2028 .
Meta allegedly used employee activity to train AI models
A leaked Meta all-hands recording says the company tracked employee Gmail, Google Chat, Metamate, and VS Code activity as part of its Model Capability Initiative. The revelations surfaced alongside layoffs affecting roughly 8,000 employees, intensifying debate about AI training consent and workplace surveillance .
Trump White House scraps AI safety executive order
The White House canceled an AI safety executive order that would have created a voluntary 90-day pre-launch review for frontier models with NSA involvement in classified testing. The decision reportedly came after direct calls from Zuckerberg, Elon Musk, and David Sacks .
Google I/O 2026 puts Gemini 3.5 at the center of its AI strategy
Google announced Gemini 3.5 as a new core architecture and made Gemini 3.5 Flash the default engine across consumer apps and Search. The company also highlighted Gemini Omni for multimodal generation and a continuously running personal agent called Gemini Spark .
Google Search becomes a multimodal AI interface
Google is updating Search so Gemini 3.5 Flash powers results globally, with the search box accepting text, images, files, videos, and active Chrome tabs. This positions Search as an AI-native interface rather than a traditional keyword tool .
Google launches Gemini for Science
Google introduced Gemini for Science, a research-focused suite that includes an experimental hypothesis-generation tool using multiple agents to debate and evaluate ideas. The move underscores growing competition to build AI systems that can assist scientific discovery, not just content generation .
Google unveils Gemini Spark as an always-on personal agent
Gemini Spark is designed to operate continuously in the background as a personal AI agent. Google’s push reflects a broader industry shift toward proactive, agentic systems that can manage tasks with minimal human prompting .
Google debuts new image creation and editing tools
Google launched Google Pics, an image generation and editing product built on the Nano Banana model. The release adds to the company’s expanding creator tools as AI image workflows become more tightly integrated into consumer products .
Google adds Managed Agents to the Gemini API
For developers, Google introduced Managed Agents in the Gemini API, signaling a stronger focus on building agent infrastructure for enterprise and app developers. This should make it easier to deploy task-oriented AI systems without assembling all the orchestration pieces manually .
NVIDIA posts record earnings and expands capital returns
NVIDIA reported record first-quarter earnings and approved an $80 billion stock buyback program. The company also raised its quarterly dividend, reinforcing how central AI-chip demand remains to its financial performance .
SpaceX files for a June 2026 IPO as AI-sector listings accelerate
SpaceX has filed a registration statement for a June 2026 initial public offering. The filing arrives amid a broader wave of high-profile tech and AI-market activity, with investors watching how capital markets price frontier infrastructure and AI-adjacent growth .