The World's Biggest AI Companies Just Failed Their Safety Report Card

Claude or Gemini? OpenAI or Anthropic? GPT-7 before year-end or not? If you read tech newsletters, you've already formed opinions on all of it.

Kalshi has real-money markets on which AI model leads benchmarks this week, which lab ships AGI first, when Anthropic releases Mythos, whether OpenAI raises ChatGPT pricing, and which company has the best coding model at year-end. These aren't abstract questions — they're live markets with real money on both sides, moving as labs ship, benchmarks drop, and announcements land.

The edge belongs to whoever actually follows this space. Not the casual observer — the person who reads model cards, tracks evals, and notices when a new release outperforms the field before the mainstream press catches up.

That person has a genuine edge. If that's you, Kalshi lets you act on it.

Trade the model race

Artificial intelligence has advanced at an astonishing pace over the past few years. AI systems can now write software, generate lifelike images, analyze complex documents, assist doctors, and even perform scientific research. Every few months, leading technology companies unveil even more powerful models, each promising to reshape the way people work and live.

But while AI capabilities continue to improve rapidly, a new independent assessment suggests that the industry's commitment to safety may not be keeping pace.

In one of the most closely watched evaluations of frontier AI developers, the world's leading AI laboratories received surprisingly low marks for how they manage the risks associated with their most advanced models. Even the highest-rated company earned only a C+, while several other well-known AI firms scored significantly lower.

The findings have sparked fresh debate over whether the race to build smarter AI is moving faster than efforts to make it safe.

The report evaluated major AI companies on a wide range of safety practices rather than on the intelligence of their models. Instead of asking which chatbot is the smartest, researchers examined whether companies have the policies, governance, transparency, and risk management systems needed to develop increasingly powerful AI responsibly.

The results surprised many observers.

Despite investing billions of dollars in AI research and publicly emphasizing responsible development, none of the leading laboratories received an "A" grade. The best-performing organization achieved only a C+, indicating that while some safety measures are in place, significant weaknesses remain.

Several other prominent AI developers received even lower scores, highlighting substantial gaps in areas such as governance, external oversight, transparency, and preparedness for increasingly capable AI systems.

For an industry developing technologies that could influence healthcare, education, finance, national security, and scientific research, the report serves as a sobering reminder that technical progress alone is not enough.

Today's AI models are far more capable than the systems available just a few years ago.

Modern AI can analyze vast amounts of information, generate convincing text, create realistic images and videos, write sophisticated software, and increasingly perform complex reasoning tasks. Companies are already integrating these systems into workplaces, government services, and critical infrastructure.

As AI becomes more powerful, the consequences of mistakes also become more significant.

An inaccurate recommendation in a casual chatbot may be inconvenient. However, errors in medical diagnosis, financial decision-making, cybersecurity, or autonomous systems could have far more serious consequences.

Safety is therefore no longer a theoretical concern. It has become an essential part of deploying AI responsibly.

Rather than measuring model performance, the assessment focused on the systems companies have in place to reduce risks.

Researchers examined whether organizations conduct rigorous safety testing before releasing new models, maintain clear governance structures, publish transparent information about risks, prepare emergency response plans, and invest in long-term AI safety research.

The report also considered whether companies allow meaningful external scrutiny instead of relying solely on internal evaluations.

Strong AI safety is not simply about preventing software bugs. It involves ensuring that increasingly capable systems remain reliable, secure, resistant to misuse, and aligned with human intentions.

According to the assessment, many companies still have considerable work to do.

One reason safety remains challenging is the extraordinary pace of competition.

The AI industry has become one of the most competitive sectors in technology. Companies such as OpenAI, Anthropic, Google DeepMind, Meta, Microsoft, xAI, and others are investing billions of dollars to develop larger and more capable models.

Each major breakthrough creates pressure for competitors to respond quickly.

This intense competition encourages rapid innovation, but it can also reduce the time available for extensive testing, independent review, and careful evaluation before new systems reach the public.

Industry leaders have repeatedly acknowledged this tension. Building safer AI often requires moving more cautiously, while competitive markets reward faster releases.

Balancing these competing priorities has become one of the industry's greatest challenges.

The report arrives at a time when governments around the world are increasing their scrutiny of AI development.

Regulators in the European Union, the United States, the United Kingdom, and several Asian countries are exploring new rules governing transparency, accountability, security testing, and deployment of advanced AI systems.

Rather than focusing only on innovation, policymakers are asking whether companies have sufficient safeguards to manage technologies that could affect millions—or even billions—of people.

Independent safety assessments like this one are likely to play an increasingly important role in informing future regulation.

Interestingly, the report suggests that the next phase of AI competition may not be determined solely by who builds the smartest model.

As AI becomes embedded in hospitals, banks, schools, governments, and large enterprises, customers will increasingly ask a different question:

Can this system be trusted?

Organizations adopting AI need confidence that models behave predictably, protect sensitive data, resist malicious misuse, and remain reliable under challenging conditions.

Companies that demonstrate strong governance and transparent safety practices may gain a significant competitive advantage, even if their models are only marginally less powerful than those of rivals.

In other words, trust could become just as valuable as intelligence.

Receiving a C+ may not sound disastrous in an ordinary classroom, but for companies developing technologies with potentially global impact, it is a warning that expectations are rising.

The report does not suggest that AI companies are acting irresponsibly across the board. Many have invested heavily in alignment research, red-teaming exercises, safety frameworks, and dedicated governance teams. However, the assessment indicates that these efforts still fall short of what many independent experts believe is necessary as AI systems grow increasingly capable.

The findings also reinforce a broader truth about artificial intelligence: progress should not be measured only by faster models, larger data centers, or higher benchmark scores.

The future of AI will depend not only on what these systems can do, but on whether society can trust them to do it safely.

As the race toward more powerful artificial intelligence accelerates, the companies that invest most seriously in safety, transparency, and accountability may ultimately prove to be the true leaders of the AI era. After all, building smarter machines is an extraordinary achievement—but building machines that people can trust may be an even greater one.