AI Safety Abandoned: What Happens Next?

ethics history

The End of the AI Safety Debate

For years, researchers, ethicists, and policymakers warned about the potential dangers of unchecked artificial intelligence development. They debated p(doom) probabilities, AI alignment strategies, and regulations to prevent catastrophe. But now, that conversation has all but collapsed.

Frontier AI companies—OpenAI, Anthropic, Google DeepMind, and others—have shifted their focus. They are no longer discussing pausing AI progress or evaluating existential risks. Instead, they are racing to deploy increasingly advanced models with a singular goal: dominance. AI safety, once a core issue, has become little more than a PR footnote.

The Cost of AI is Dropping to Near Zero

The rapidly declining cost of both AI training and inference is often overlooked. Training state-of-the-art AI once required billions of dollars, but today, open-source models can be fine-tuned on consumer GPUs at a fraction of the cost.

Additionally, API-based access to powerful AI systems is getting cheaper. What once cost thousands now costs pennies—or nothing at all. Companies are slashing prices to stay competitive, while innovations in efficiency (like better quantization and hardware acceleration) make AI more accessible than ever.

China’s AI Revolution Despite GPU Restrictions

The U.S. has imposed export restrictions on advanced AI chips like NVIDIA’s A100 and H100, aiming to slow China’s AI advancements. This strategy has failed.

Chinese labs and companies are building and training capable AI models using less powerful GPUs, optimizing software, and ramping up domestic chip production. While still behind NVIDIA in raw performance, China is closing the gap quickly. AI development is now a truly global arms race.

Bad Actors and AI-Powered Scam Bots

Cybercriminals are leveraging AI in sophisticated ways:

  • Automated scam bots can impersonate real people and socially engineer victims far more effectively than traditional phishing attacks.
  • AI-generated fraud includes deepfake videos, voice synthesis, and hyper-realistic fake identities.
  • AI-enhanced hacking tools automate reconnaissance, exploit discovery, and attack execution at unprecedented speeds.

As AI costs drop, these tools become accessible to lower-level criminals, not just nation-states. Defenses against AI-driven threats are struggling to keep up.

The Job Market is Collapsing Under AI-Generated Content

The idea that AI would “augment” rather than replace human work is quickly falling apart. AI-generated content is making many traditional roles obsolete, including:

  • Copywriting and journalism: AI can generate engaging articles at scale.
  • Graphic design and illustration: AI tools produce high-quality visuals in seconds.
  • Customer service: AI chatbots are replacing human support agents.
  • Video production: AI-generated video content is improving rapidly.

Businesses are embracing AI-driven automation, and the promise of “new jobs replacing the old ones” is looking increasingly shaky.

So, What Happens Now?

The genie is out of the bottle, and AI safety is no longer a priority. Governments are playing catch-up, bad actors are exploiting AI, and industries are being transformed in real time.

Possible Future Scenarios

  • Regulatory Crackdowns (Too Little, Too Late?): Governments may introduce AI regulations, but enforcement will be challenging as AI development continues at breakneck speed.
  • The AI Bubble Bursts (Or It Doesn’t): AI could hit diminishing returns, but even if progress slows, the technology is already transformative.
  • Acceleration to AGI (And Then?): Companies are racing toward artificial general intelligence (AGI). If they succeed, it could reshape civilization in unpredictable ways.

AI safety is dead as a mainstream concern. What remains is an AI arms race with little oversight and rapidly declining costs. We are in uncharted territory.

Is AI Just a Glorified Autocomplete?

Some researchers believe AI is limited by its fundamental nature:

  • Lack of True Understanding: AI models don’t “know” anything—they generate statistically likely responses.
  • Surface-Level Reasoning: AI lacks deep comprehension, common sense, and self-awareness.
  • Failure Cases: AI confidently generates incorrect information without a way to verify truth.

AI Progress Requires Fundamental Breakthroughs

AI’s improvements come from scaling up models, refining algorithms, and enhancing hardware. However, without breakthroughs in cognitive modeling, AI may remain impressive but not revolutionary.

The Human Brain’s Unique Strengths

  • Efficiency: The human brain operates on just 20 watts, while AI models consume megawatts.
  • Generalization: Humans adapt easily to new challenges, unlike AI.
  • Creativity: AI can generate novel outputs but lacks true originality.
  • Common Sense: Humans naturally infer meaning and navigate ambiguity.

AI Progress May Be Net Positive

If AI is fundamentally limited, rapid progress may be more about automation than existential risk. This suggests:

  • AI will augment rather than replace human intelligence.
  • Concerns about AGI may be overstated.
  • Focus should shift to AI’s societal integration rather than existential fears.

If AI researchers see limits to its capabilities, then breakthroughs are necessary to match human intelligence. The question remains: Are those breakthroughs possible in the foreseeable future?

What Do You Think?

Will AI continue its rapid progress, or are we already approaching a plateau?