When Sam Altman first told ABC News in March 2023 that he and his OpenAI colleagues were “a little bit scared” of just how good the artificial intelligence (AI) systems they were unleashing are, it sounded like a cautious scientist’s plea for sober guardrails. Two years on, the comment has aged into a mission statement: build fast, warn loudly, and hope the world keeps up.

From Cautionary Quote to Capitol Hill

Altman’s wariness is now a regular talking point in Washington, D.C. In May, he appeared before the Senate Commerce Committee alongside Microsoft (MSFT) and Advanced Micro Devices (AMD) executives, urging lawmakers to craft rules that let the United States “win the [AI] race” while protecting society from deepfakes and cyberattacks. 

“I believe this will be at least as big as the internet—maybe bigger,” Altman told senators, framing regulation as a national‑security imperative.

Just weeks earlier, at TED 2025 in Vancouver, Altman confessed that OpenAI’s GPUs are “melting” under demand and that the company’s 800 million weekly users have left its engineers “exhausted and stressed.”  The tense 47‑minute interview covered topics ranging from artist compensation to autonomous “agentic” AIs that could roam the web on a user’s behalf — “the most consequential safety challenge” OpenAI has faced, Altman admitted.

Don't Miss

• Think it’s too late to invest in the booming AI sector? This one’s still under the radar

• How This Startup is Disrupting the $4.83 Trillion Wealth Engine That’s Been Hidden From Everyday Investors for Over 93 Years

New AI Models Keep Dropping

Despite the CEO’s qualms, the AI model release cadence is only accelerating. On May 14, OpenAI rolled out GPT‑4.1 and its lighter sibling GPT‑4.1 mini to ChatGPT subscribers, touting sharper coding skills and stricter instruction‑following than last year’s GPT‑4o. The update followed April’s debut of the heavyweight o3 model and a research preview of GPT‑4.5, underscoring how quickly each “next” model is eclipsed.

The breakneck progress is colliding with a brutal labor market reset. According to industry tracker Layoffs fyi, more than 61,000 tech workers have already lost their jobs in 2025. Amazon (AMZN), Google (GOOGL)(GOOG), Meta (META), and others trimmed headcount this month, while Microsoft executed its largest cut since 2023 — about 6,000 roles, or 3% of staff — as it funnels $80 billion into AI infrastructure.

At Google, more than 1,300 employees have signed a petition demanding voluntary buy‑outs and fairer severance amid rumors of another AI‑driven reorg, highlighting growing anxiety inside companies that are simultaneously evangelizing and automating.

Altman’s original quote lands differently in 2025: the fear inside OpenAI is no longer just about hypothetical misuse; it’s about GPU shortages, investor politics, congressional scrutiny, and a job market reshaped by the very tools his company sells. Yet the CEO insists a dose of dread is healthy, and proof that creators still feel the weight of the technology they steer.

If the past two years are any guide, that tension between exhilaration and apprehension will only intensify as ever‑smarter models roll out — whether or not the rest of society is ready.