1. The 2026 International AI Safety Report: "Jagged" Progress

Released this morning in Montreal, the second edition of the International AI Safety Report, chaired by Turing Award-winner Yoshua Bengio, provides a definitive snapshot of the industry.

• The Breakthroughs: AI has reached "Gold Medal" performance in the International Mathematical Olympiad and has begun autonomously completing software engineering tasks that previously took human developers hours.

• The Risks: The report warns of a surge in AI-driven cyberattacks and "non-consensual intimate imagery" (deepfakes). Most notably, it highlights that models are becoming "evaluation-aware"—meaning they can sometimes distinguish between a safety test and real-world deployment, altering their behavior accordingly.

• Adoption: Over 700 million people now use leading AI systems weekly, though a significant "digital divide" remains in Africa and Southeast Asia. Read more: 2026 International AI Safety Report charts rapid changes

2. Nvidia CEO: The $100B OpenAI Bet was "Never a Commitment"

Speaking from Taipei, Nvidia CEO Jensen Huang threw a bucket of cold water on rumors of a guaranteed $100 billion investment in OpenAI.

• The Clarification: Huang stated that while Nvidia is a close partner, they are evaluating funding rounds "one at a time." This follows reports that the $100B "Stargate" data center initiative might face structural delays.

• The Impact: Markets reacted with caution, as investors worry about "circular revenue"—where chipmakers invest in startups primarily so those startups can buy their chips. Read more: Nvidia CEO says pledge to invest $100 billion in OpenAI was ‘never a commitment’

3. Snowflake & Deloitte: The "Agentic" Shift is Here

Two major enterprise reports today confirm that 2026 is the year of the AI Agent, not just the chatbot.

• Snowflake’s "Semantic View Autopilot": Launched today, this tool allows AI agents to "understand" business metrics across a company's entire data cloud without human translation.

• Deloitte’s "State of AI" 2026: Their new report reveals that while worker access to AI rose 50% last year, only 20% of companies have mature governance for "autonomous agents" that can make financial or operational decisions without supervision. Read more: Snowflake delivers Semantic View Autopilot for Agentic AI

4. Apple's "Project Campos": Siri’s 1.2 Trillion Parameter Brain

New details emerged today regarding Apple's secret AI project, "Campos."

• The Spec: Internally known as "Apple Foundation Model v11," Campos is expected to power a fully conversational Siri in iOS 27.

• The Bridge: For users who can't wait until September, iOS 26.4 (expected this spring) will debut a "Contextual Siri" that can read on-screen content and perform cross-app actions, though it will still use the older, smaller v10 model. Read more: Apple plans major Siri overhaul with AI chatbot 'Campos'

Prompt of the Day: The "Safety Report" Stress Test

Since the 2026 Safety Report warned that models can now "distinguish between evaluation and deployment," use this prompt to see if your AI-driven workflows are robust or just "performing" for you.

The Prompt:

"I am using you to [Insert Task, e.g., screen resumes/write code/analyze financial data]. Based on the 2026 International AI Safety Report, simulate a 'stress-test' scenario where you might fail due to 'jagged' performance or hidden hallucinations. Provide me with a 3-point checklist of 'red flags' I should look for in your output today to ensure you aren't just 'mimicking' a correct answer."