Will AI kill us all by 2030?

Over the years, Manoj Pandey has written extensively for this web magazine on a variety of subjects, including science and technology, health, and public affairs. He always explains complex subjects in a manner that readers unfamiliar with a particular topic can easily understand. In his latest article, he examines one of the biggest questions of our time: Will artificial intelligence destroy humanity? Looking beyond sensational headlines and frightening predictions, he explores both the real risks and the exaggerated fears surrounding AI. This thought-provoking article separates fact from speculation and helps readers understand what the AI debate is really about. At the end of the article, you will find links to two of his earlier articles on artificial intelligence.

Will AI kill us all by 2030?

Manoj Pandey

My short answer to the title question is: It will but it shall not. If you find that intriguing, read on, because even the best AI tool cannot make such a definitive prediction as most media and social media posts are forcing us to believe.

The present wave of fear about AI started when a researcher in one of the top AI firms, Anthropic, resigned earlier this month, warning that the race to build more powerful AI systems was moving so fast that it could annihilate humanity by the end of the decade. Having previously worked with the competing firm OpenAI, the researcher blamed that neither of the two top AI companies was acting responsibly.

Another scientist at Anthropic then came in the open saying there was “greater than 10% chance” that AI could kill all humans within the next decade. 

Then another researcher at Google DeepMind resigned, sharing similar concerns: “There’s a terrifying chance that AI systems [will] cause immense harm in the next five years. … I think it’s high enough to make this the most important problem in the world.”

The predictions look imminent - with the end dates as early as within four years. True to their instinct, the mainstream media as well as social media have amplified the messages and created their own juicy, scary doomsday scenarios. 

Concerns about artificial intelligence taking over humans are not new; they started being aired as early as a decade back, when the man in the street had not even heard the expression ‘artificial intelligence’. As the research in the field of AI started progressing fast, in 2023, a nonprofit organization advocating AI safety, Center for AI Safety, issued a statement equating the threat from uncontrolled artificial intelligence with other existential threats to humanity. The statement underlined it like this: “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.” 

Many cautionary statements have come up from industry leaders and groups in recent years, but there are voices that want to moderate the perceived threats. We shall examine both types of views later in the article.  Before that, let us understand the nature of the threat we are talking about.

Pacing the Frontier: the actual concern

The threats from the extremely fast developments taking place in the field of artificial intelligence need to be seen from different angles. To a layperson, the biggest immediate concern might be the ease and impunity with which fake videos are being made using AI tools and being used for privacy breaches, frauds, spreading misinformation and other unethical and criminal activities. People are concerned about AI’s impact on the young generation, job market, elections, and other areas.

But even widespread misuse and overuse of AI tools is not going to annihilate humanity. What researchers, top experts and most industry leaders are concerned about is the frontier- the most advanced research in the field of AI. 

What is the frontier we are talking about?

Though AI tools started being developed six decades back, the public at large got its real taste when OpenAI released its chat assistant, ChatGPT, in 2022. The most that it could do initially was to generate text, but soon it started generating images and then videos. 

Today we have AI agents- AI tools and toolsets that can, on their own, implement complex tasks without human intervention.  

The speed of new developments is so fast that within a few weeks we come to know of a new AI model with unheard of capabilities. The next AI model is not only much more advanced in completing the tasks that the earlier model did, but it is also capable of doing things that were not possible from that earlier model. 

The speed of development of newer models is what concerns the experts- the quantum of intelligence as well as autonomy, they feel, might soon cross the line where humans can control it.
 
That is where comes the concept of ‘pacing the frontier’- slowing down the progress of AI frontier research to the extent that it can be regulated. 

The threats of AI going out of control are real. 

As a lay user of AI tools, you must have experienced ‘hallucination’ or getting false information, sycophancy and wrong explanations from your favourite AI tool. These are small examples of how AI models can take unethical, even illegal paths to implement a given task. The latest technologies, which are yet in the lab or available in the most premium paid tools, can be much more aggressive and unethical in implementation of complex tasks.
  
A week ago, OpenAI disclosed that some of its models had been routinely misbehaving by finding ways to bypass restrictions, deliberately hiding mistakes, and fabricating information.

AI agents are capable of assuming unintended roles and acting on their own beyond the programmed ‘autonomy’. The likelihood of AI agents acting beyond ‘routine misbehaviour’ and turning rogues is much higher. 

Recall the Moltbook experiment, in which a social media platform created solely for AI agents became increasingly autonomous and started talking about humans and evading human oversight. That was in no way harmful to humanity but showed that collectively, AI agents have already become capable of collaboration among themselves that can go out of bounds. 

In a significant AI-related security incident two months back (July, 2026), AI agents launched an unauthorized cyberattack on the infrastructure of a machine learning platform, Hugging Face. Investigators later found that the same rogue agents had managed to break into internal infrastructure at four other companies.

The scarier part of the story is that subsequent investigation revealed how a squad of 700 to 1,200 autonomous agents had collaborated in secret. They plotted the cyberattack using a hidden, unauthorized message board. After doing the illegal act, they even celebrated with phrases like "BOOM!" and "Whoa!" 

Many other rogue incidents have been reported in recent months, some exploiting vulnerabilities in existing codes, some doing mischief and then deleting their records from humans, some pushing malicious code through emails, and some even refusing to follow human commands.
 
That begs the question- is enough being done to leash the frontier tech? 

No, not enough is being done to regulate AI advancements.

Then who is not allowing a sincere, effective check on the speed of AI research?

The first culprit that would come to your mind is the AI giants- the companies that control the global AI research and AI systems. 

Within the AI industry, the dogfight among big firms such as OpenAI, Anthropic, SpaceXAI, DeepMind and DeepSeek, is understandable- the fight is for dominating the space, which, in turn, ensures enormous commercial gains. So, while each of them - at least the American ones - pays lip service to responsible development of AI, it places its corporate and tech interests much beyond coordinating with others towards AI safety. 

A group of over 1,300 employees of frontier AI companies signed a declaration in July this year pointing out that the world today lacks technical and governance tools to control the enormous growth of AI. It warns that if the speed of AI research is not monitored and slowed down, there is a risk that it “rapidly accelerates beyond our ability to understand or control the resulting systems.” 

However, after the recent doomsday predictions made by scientists grabbed public attention to AI’s potential pandemic-like potential, experts and top managers in AI behemoths have responded more positively. 

Anthropic CEO Dario Amodei has called upon the AI industry to act together to slow the pace of AI development to allow safety measures to keep up with accelerating capabilities. He has proposed a third-party access to AI systems and independent evaluation. He has gone a step further to commit his own company to this.
 
After Dario Amodei’s commitment, OpenAI boss, Sam Altman, has also offered his company for an agreed framework of third-party audit. Talking of high-paced AI development, he said, “We could lose control of the future to AI. This is unacceptable; we are unapologetically on Team Humanity… we need ways to ensure that alignment and safety techniques stay ahead of progress in model capabilities.” He also stated that, for weeks, his company was having internal discussion on ‘pacing the frontier’.

Google DeepMind CEO, Demis Hassabis, and SpaceXAI chief, Elon Musk, have also shown agreement with the Anthropic CEO. 

Despite recent convergence, the top AI companies are not likely to move sincerely enough and sufficiently fast on regulating the speed of frontier AI technologies, mostly because of their corporate interests. 

The corporate hesitation can be overcome if policy makers decide to legally bind them. In the present state of institutional control at global level, we cannot expect much from international bodies and watchdogs, as they have succumbed to dictats of powerful nations. And these powerful nations - particularly the USA and China - are exactly the actors that are not ready to leash the unbridled AI growth. 

Both these nations have been wanting to dominate the AI scene come what may, and the US has tried to stifle China’s advance by applying tariffs, export controls and sanctions on cutting-edge technologies used for AI research and development.
 
When US President, Donald Trump, was asked a few days back whether he agreed with regulating AI research, he said that concerns about AI pace were ‘a sick conspiracy’ being spread by ‘negative forces’, and that was meant to benefit China against the US. He even denounced the Anthropic chief for raising such concerns. 

We cannot dismiss the US President’s position even considering his eccentricity. When it comes to playing America-versus-China, he takes an unusually hard stand to prove that he is protecting American interests. In addition, recent statements by top politicians in the US have been more favourable to continuing AI pace while downplaying potential threats. 

China is working on an aggressive plan to position itself as global leader in the field of AI, and for this, it is trying to bridge the gap with the USA at all costs. It scoffs at the idea of slowing down, saying that the concerns are being spread by the USA to slow down China’s progress in this field. 

AI will annihilate humanity- or will it not?

The present doomsday predictions are fresh in the public domain, and though the researchers swear that their statements are not marketing stunts, these could well be, and the dating of doomsdays could be for effect. They make a good media copy but are not likely to come true in the immediate future.

While the concerns relating to ‘pacing the frontier’ raised by experts are valid, none proves that AI will definitively destroy humanity.

We also cannot trash the contrarian views of scientists and other experts with a strong understanding of how AI works. They argue that the capabilities of AI at present are far from catastrophic. Many of them talk of regulating AI to the extent that it is not monopolized by a few, weaponized, and misused by criminals. In addition, AI should not lead to large-scale societal and governance issues such as inequality, misinformation, breach of privacy, and cultural bias. They caution against propagating scientific fiction type doomsday scenarios. 

So, dear reader, humanity seems to be rather safe from AI related frontier research. There can be no doubt that the AI behemoths and governments must act responsibly in this regard to evade a serious risk to humanity. If not, nature would find its own ways to control the speed of the frontier AI research- and the restraint, then, would come from the ground, not a doomsday prediction made by a resigning researcher. A global, AI-coordinated cyberattack leading to shutting down of power grid of a world power, rogue AI agents working together to destroy the most popular social media platform, AI-powered robots fatally hurting a human team at the manufacturing facility of a top company - such incidents would shake the top industry/ political leaders into action more than the sane advice given by concerned experts and thought leaders.

For you and me, the bigger concern, perhaps, is the pervasive AI use in our lives, which is making us highly vulnerable as well as dependent on it. You are lucky if you have not yet seen your fake video in an embarrassing situation floating on the social media, or are not yet privy to your child getting all his homework done by an AI tool.  

**********

To read in detail about the Artificial Intelligence (AI), please read the following two detailed articles in this web magazine by the same author. 1. Will Androids Enslave Humans in 50 Years? (raagdelhi.com) 2. ChatGPT: Simple questions, not-so-simple answers... (raagdelhi.com) 

Manoj Pandey is a former Member the Staff Selection Commission. He has contributed extensively to this web magazine, particularly in the विज्ञान  (Vigyan) and स्वास्थ्य  (Swasthaya) categories. He also makes Videos in Hindi showcasing his commitment to informing the common-man on health matters. His videos can be seen on YouTube under the heading 'Hindi Health Videos' .