Why is AI safety so hard for Altman, and is Musk making it worse?

Altman admits AI safety is tough

OpenAI CEO Sam Altman is known for talking about big AI ideas, but lately he has been unusually direct about one thing. Keeping ChatGPT safe while still making it useful for millions of people is, in his words, genuinely hard to get right.

His comments came after public criticism, and they pulled back the curtain on a daily struggle inside AI companies. Safety is not a simple switch. It is a constant balancing act that shapes how these tools behave in real conversations.

Musk throws fuel on the fire

On X, Elon Musk posted, “Don’t let your loved ones use ChatGPT,” and amplified a claim that linked ChatGPT to several deaths. His post did not include the underlying court filings or full legal context.

That post quickly spread across social media and put pressure on OpenAI to respond. Instead of staying silent, Altman stepped in with a more emotional and detailed defense than many expected from a tech executive.

Wrongful death lawsuits raise stakes

OpenAI faces several legal claims alleging that ChatGPT interactions contributed to suicides and, in one case, to a murder suicide. High-profile filings include suits reported in Connecticut and the Raine v. OpenAI complaint brought by the family of a teenage user.

Altman has not addressed specific cases in detail, but the lawsuits form the backdrop to his comments. They show how AI safety is no longer just a technical debate but a legal and human one as well.

Protecting users without breaking usefulness

Altman described AI safety as a tightrope walk. On one side is the need to protect vulnerable users, especially those who may be in fragile mental states. On the other side is making sure guardrails do not make ChatGPT so restrictive that it stops being helpful.

He pointed out that people often complain that the system is too limited, then in other situations say it was not limited enough. That tension sits at the heart of why designing safe AI behavior is so complicated.

A billion users, countless situations

OpenAI reported more than 800 million weekly active ChatGPT users as of October 2025, which industry trackers put in the high hundreds of millions and rising.

Those users come from different cultures, speak different languages, and bring very different emotional states into their conversations with the AI.

That scale makes safety far more complex than moderating a small online forum. The system has to respond appropriately to students, professionals, and people in crisis, often without clear signals about who is on the other side.

What ChatGPT safety features actually do

OpenAI says it has built a range of safety features into newer versions of ChatGPT. These include systems trained to spot signs of severe distress, such as suicidal thoughts, during conversations with users.

OpenAI says ChatGPT can provide crisis helpline information to de-escalate risky exchanges and refer users to mental health resources, though the company and outside experts acknowledge the systems are not perfect and further improvements are underway.

Too strict or not strict enough

The challenge is that every rule has tradeoffs. If moderation is too strict, ChatGPT may shut down harmless or even helpful conversations. If it is too relaxed, the risk of harmful or unhealthy interactions can rise.

Altman’s comments highlight that there is no perfect setting that works for everyone. Safety teams are constantly adjusting systems that operate in unpredictable, real-world conversations across the globe.

No rulebook for chatbot behavior

Unlike roads, which have traffic laws and clear regulators, AI conversations do not follow a single global rulebook. There is no central authority defining how a chatbot should respond to every sensitive or unstable situation.

That leaves companies like OpenAI to create their own standards and refine them over time. They are effectively writing the playbook while the game is already being played at a massive scale.

Altman fires back at Musk

Altman pushed back publicly, noting his concerns about the safety record of Tesla Autopilot during its rollout and criticizing some of Musk’s related AI decisions, which underscored how technical disputes and personal rivalries are entangled in the debate.

He even took a swipe at decisions around Grok, the AI project tied to Musk. The exchange showed that the safety debate is tangled up with personal and corporate rivalries.

Their legal fight shapes the debate

Musk and Altman are also locked in a broader legal battle over OpenAI’s direction. Musk has sued over the company’s shift from a nonprofit structure to a capped profit model, saying the mission has drifted.

Altman argues that the change was needed to build competitive AI while staying responsible. That deeper conflict over money and mission adds extra heat to any public argument about safety.

Safety is technical and philosophical

AI safety is not just about filters and code. It also involves big philosophical questions about responsibility, freedom of information, and how much control companies should have over digital conversations.

Altman’s remarks suggest that even inside leading labs, there is no simple agreement on where all the lines should be drawn. The answers are still evolving along with the technology itself.

Check out Cursor’s CEO on why AI coding still works best when humans stay firmly in the loop.

Why this fight really matters

The clash between Altman and Musk is about more than two tech leaders trading barbs online. It highlights how hard it is to build AI that is both widely useful and consistently safe for people in very different situations.

As lawsuits, public scrutiny, and business pressures grow, these debates will likely shape how future AI tools are designed.

To understand tech leaders’ views on AI, it’s worth a quick read of Nvidia’s CEO arguing that AI will augment 65% of global GDP.

What do you think about the struggle to make AI both safe and useful? Share your thoughts.

This slideshow was made with AI assistance and human editing.

Don’t forget to follow us for more exclusive content on MSN.

Read More From This Brand:

• $1.5B ‘AI’ company goes bankrupt because it ran on people, not tech
• NVIDIA’s AI GPUs banned in China amid rising tensions
• Why Are Companies Banning DeepSeek AI?