OpenAI CEO Sam Altman has described GPT-5 as the most powerful artificial intelligence model ever developed by the company — and, surprisingly, admitted that it even left him “a little scared.”

This revelation came during a podcast interview where Altman discussed some insights into the soon-to-be-launched AI model, which OpenAI plans to unveil in August 2025. While details about the model’s full capabilities remain under wraps, Altman shared enough to raise eyebrows in the tech community and beyond.

A Leap Beyond Expectations

According to Altman, GPT-5 feels significantly faster and more intelligent than any of its predecessors. He emphasized that in some private testing sessions, the model demonstrated abilities so advanced and unpredictable that they made him genuinely uneasy.

“I’ve worked on AI for years,” Altman said, “but during certain sessions with GPT-5, it felt like the model understood things at a level that was deeply surprising — and sometimes frightening.”

While he didn’t disclose any technical specifics or precise capabilities, his comments suggest GPT-5 may represent a dramatic leap in how AI models process information, interact with humans, or generate original content.

Comparing GPT-5 to the Manhattan Project

Perhaps the most shocking part of Altman’s interview was his comparison of GPT-5 to the Manhattan Project — the secret U.S. government initiative during World War II that led to the development of the atomic bomb.

By invoking this analogy, Altman didn’t imply that GPT-5 was a weapon, but he highlighted the immense power and potential consequences of such transformative technology. Just as nuclear physics revolutionized the world in the 20th century, advanced AI could redefine society, governance, economics, and even personal identity in the 21st.

The comparison was meant to underline that GPT-5 might not just be a technological advancement, but a civilizational turning point — one that comes with both incredible promise and significant peril.

The Fear Factor: What’s Causing Concern?

Although Altman refrained from specifying exactly what frightened him about GPT-5, there are several areas where such a model might raise serious ethical and security alarms:

1. Unpredictable Reasoning: If GPT-5 can reason in ways humans don’t fully understand, it could create outputs or decisions beyond human control or comprehension.
2. Persuasive Abilities: A more advanced model might convincingly imitate real people or manipulate public opinion at scale.
3. Autonomous Decision-Making: GPT-5 may be capable of making or suggesting decisions with minimal human oversight in complex domains like medicine, law, or warfare.
4. General Intelligence Signals: If GPT-5 begins to show signs of Artificial General Intelligence (AGI) — the theoretical point where AI can perform any intellectual task a human can — this would radically shift humanity’s relationship with machines.

Regulation Lags Behind Innovation

Altman also voiced a long-standing concern: the lack of effective global regulation for artificial intelligence. He pointed out that while AI is advancing at a rapid pace, governance structures are still nearly non-existent or inadequately equipped to deal with the risks.

“There are almost no enforceable laws on AI right now,” he said, “and that’s dangerous.”

He has previously warned governments and tech leaders that without proactive global frameworks, advanced AI could be misused — from mass surveillance to deepfakes, disinformation campaigns, or even autonomous weapons.

Not the First Time Altman Has Sounded the Alarm

This isn’t the first instance where Altman has expressed apprehension about the potential misuse of AI. He has often emphasized the ethical responsibility that comes with building powerful models.

In earlier interviews, he has stressed that OpenAI must proceed with caution, ensuring that safety, alignment, and societal well-being are prioritized at every step of development. He’s also been a vocal advocate for “alignment research” — the scientific discipline focused on ensuring AI systems remain loyal to human values.

GPT-5: What We Know So Far

Though technical details are scarce, leaked information and expert speculation suggest that GPT-5 may offer:

• Significant improvements in contextual reasoning
• Longer memory spans, possibly enabling conversations or workflows over days or even weeks
• Multimodal capabilities, seamlessly integrating text, voice, image, and even video generation
• More human-like interaction, with nuanced emotional recognition and response
• Better self-correction, allowing the model to revise its mistakes in real-time

If these features are confirmed, GPT-5 won’t just be a tool — it may become something closer to a digital assistant with deep cognitive abilities.

Balancing Innovation with Responsibility

Altman’s concerns about GPT-5 aren’t a rejection of AI’s potential — far from it. He remains an optimist about the transformative power of artificial intelligence. However, he also acknowledges that transformative power comes with immense responsibility.

“We have to build systems that are beneficial to humanity,” he said, “and make sure we never lose sight of the consequences, especially when these systems get smarter than we expect.”

As OpenAI prepares to launch GPT-5 in August 2025, the world watches with anticipation — and caution. This new model may redefine what’s possible in computing, communication, creativity, and decision-making.

But Altman’s remarks serve as a reminder: in the race to create intelligent machines, the ultimate challenge is not just technological — it’s ethical, societal, and existential.