New York’s RAISE Act Puts OpenAI and Anthropic Under New AI Safety Rules

The race to build more powerful artificial intelligence models has created enormous commercial opportunities. It has also created a difficult question for governments: how do you regulate technology that can change faster than the laws designed to control it?

New York Targets AI Risks With New Rules for Chatbots and Frontier Models

New York is preparing to answer that question with one of the most significant AI safety laws in the United States. Its RAISE Act will require major AI developers, including OpenAI and Anthropic, to register with the state, publish safety information, and report serious incidents.

This is more than another compliance requirement. It signals that the business of building advanced AI is entering a new phase, where safety procedures and public accountability could become as important as model performance.

New York Opens the Door to Formal AI Oversight

New York has opened a registration portal for large developers of advanced AI systems. Companies covered by the RAISE Act will be required to register by January 1, 2027.

The law targets frontier AI developers with more than $500 million in annual revenue. These are companies building the most advanced models, rather than smaller businesses simply using existing AI tools.

The requirements include three major obligations:

• Publish safety frameworks on company websites.

• Report critical safety incidents within 72 hours.

• Release public reports explaining model capabilities and safety measures.

The state can impose penalties of up to $1 million for an initial violation and $3 million for subsequent violations.

The message is clear. New York wants large AI companies to demonstrate that they have systems for managing risks, not simply promise that their technology is safe.

DIGIT Becomes the New Gatekeeper

The law will be implemented by the Office of Digital Innovation, Governance, Integrity, and Trust, known as DIGIT.

Its role is to oversee major AI developers and enforce the requirements established by the RAISE Act. Companies that fail to register may also face enforcement action, although state officials have not provided detailed information about how those cases will be handled.

This creates a new layer of government oversight between AI developers and the markets in which they operate.

For companies such as OpenAI and Anthropic, regulatory planning will increasingly need to happen alongside product development. Safety documentation, incident response, and compliance cannot be treated as administrative tasks added after a model has launched.

They are becoming part of the operating system of an AI business.

A $55 Million Center Will Independently Evaluate AI Systems

New York is also investing in independent research.

Governor Kathy Hochul announced the New York Center for AI Responsibility and Research at Binghamton University, part of the State University of New York system.

The center was initially announced in January with $55 million in donations and capital.

Its purpose is to independently evaluate the safety, security, and operation of advanced AI systems. It will also work with DIGIT to review the safety frameworks submitted by companies.

This matters because AI developers should not be the only organizations deciding whether their own products are safe.

Independent evaluation can help governments, businesses, and the public understand the risks of advanced models. It can also create a more consistent basis for comparing the safety practices of different companies.

The challenge will be ensuring that these evaluations remain technically credible and keep pace with rapidly changing models.

The Rules Go Beyond Frontier AI Models

The RAISE Act is part of a wider effort by New York to address AI-related risks.

Last year, the state began enforcing a law requiring operators of AI companion services to regularly remind users that they are interacting with a bot. Companies must also make reasonable efforts to detect signs of suicidal thinking and self-harm.

A separate law signed in October focuses on chatbots used by minors. It restricts deceptive human-like interactions, emotionally manipulative or excessively agreeable responses, and encouragement of harmful behavior, including suicide and disordered eating.

That law is scheduled to take effect on June 1, 2027.

Together, these measures show that regulation is expanding beyond the technology itself. Governments are increasingly examining how people interact with AI, especially when users may be vulnerable or unable to distinguish between a machine and a human.

California and Illinois Are Building Their Own AI Rulebooks

New York is not acting alone. Other states are developing similar requirements, creating a more complicated regulatory environment for companies operating across the United States.

California's SB 53 took effect in January. Illinois' SB 315 is scheduled to take effect at the beginning of 2027.

The differences matter. New York requires faster reporting of critical incidents than California, while Illinois adds independent safety compliance audits.

For a company developing one model and selling it across multiple states, these differences can create additional legal, technical, and operational work.

The United States does not yet have one uniform approach to frontier AI regulation. Companies must prepare for a patchwork of state requirements.

AI Safety Is Becoming a Competitive Business Capability

For years, the competitive advantages in AI were relatively easy to describe: better models, more computing power, more data, and faster product development.

That equation is changing.

As regulation expands, companies must also demonstrate that they can identify risks, respond to incidents, protect users, and explain how their systems operate.

These requirements create costs. Companies need legal teams, safety researchers, compliance specialists, testing systems, and reliable incident-reporting processes.

But they may also create competitive advantages.

Large companies such as OpenAI and Anthropic have more resources to build dedicated safety and compliance teams. Smaller competitors may struggle to absorb the same costs.

This could strengthen the position of established AI providers by making it more expensive for new companies to compete at the frontier.

At the same time, credible safety practices could become a selling point. Businesses choosing an AI provider may increasingly ask not only which model performs best, but also which provider can demonstrate responsible operations.

Trust becomes part of the product.

The Biggest Risk Is Regulation That Cannot Keep Up

AI systems evolve quickly. A model can receive new capabilities, be integrated into new products, or reach millions of additional users through a software update.

Regulation moves more slowly.

If reporting requirements are unclear, companies may struggle to determine which incidents qualify as critical. If safety frameworks become overly standardized, they could encourage companies to focus on documentation rather than actual risk reduction.

There is also a potential cost for innovation. Compliance expenses may be manageable for major laboratories but disproportionately difficult for startups.

The challenge for New York is to create meaningful oversight without turning regulation into a barrier that only the largest companies can afford.

The effectiveness of the RAISE Act will depend not just on penalties, but on whether its rules lead to measurable improvements in safety.

What the New Rules Mean for AI in the Business World

The immediate requirements target large frontier developers, not every company using an AI chatbot or productivity tool.

However, the impact will travel through the technology supply chain.

AI providers may introduce stronger documentation, clearer product limitations, better incident reporting, and more detailed customer assurances. Enterprise buyers may demand evidence of safety testing and compliance before approving AI systems for important operations.

That could influence pricing and purchasing decisions.

A business selecting an AI platform may eventually compare providers on several dimensions: performance, cost, reliability, data protection, and regulatory readiness.

For AI companies, this creates an opportunity to turn compliance into a commercial advantage. For customers, it provides more information when deciding which systems deserve their trust.

What Small and Medium-Sized Businesses Should Do Now

Smaller businesses do not need to build their own frontier AI safety departments. But they should prepare for a market where responsible AI use becomes increasingly important.

Choose providers carefully. Ask vendors how they protect customer data, handle incidents, and explain the limitations of their models.

Document AI usage. Keep a basic record of which tools employees use, what information they process, and which business activities depend on AI.

Protect sensitive information. Avoid placing confidential customer, financial, or employee data into AI services without understanding their data policies.

Keep humans responsible for important decisions. AI can support customer service, marketing, research, and administration, but high-impact decisions need appropriate human review.

Make compliance part of procurement. When purchasing software, consider whether the provider can supply clear security, privacy, and safety information.

AI should reduce business costs and improve productivity. It should not introduce risks that the company cannot explain or manage.

The Next AI Advantage May Be Trust

New York's RAISE Act marks an important change in the economics of artificial intelligence.

The industry is moving beyond a period in which building the most capable model was the primary objective. Companies must now show that they can manage the consequences of deploying increasingly powerful systems.

OpenAI and Anthropic will face new registration, reporting, and transparency requirements. Independent research institutions will gain a larger role. Other states are developing their own approaches, adding complexity to the market.

The winners will still need strong technology. But technical performance alone may no longer be enough.

In a market where AI is becoming essential infrastructure, trust, transparency, and operational discipline could become some of the most valuable competitive advantages.

New York Targets AI Risks With New Rules for Chatbots and Frontier Models

New York RAISE Act, AI safety regulation, OpenAI regulation, Anthropic regulation, frontier AI developers, AI transparency, AI safety frameworks, DIGIT oversight, AI incident reporting, California SB 53, Illinois SB 315, AI compliance, AI governance, responsible AI