Key Points:
- OpenAI launched GPT-5.4-Cyber to help organizations spot and fix software security vulnerabilities.
- The company will roll out the new artificial intelligence model to hundreds of trusted testers before expanding to thousands of testers.
- Rival company Anthropic recently released its own powerful cybersecurity model, Mythos, to select corporate partners.
- Government officials and Wall Street leaders worry these advanced coding tools could fall into the hands of cybercriminals.
OpenAI just launched a brand new artificial intelligence tool designed specifically for cybersecurity. The creator of ChatGPT announced on Tuesday that it will allow a select group of users to access a new model called GPT-5.4-Cyber. This powerful software excels at spotting deep security vulnerabilities hidden inside complex computer code. The release comes just one week after rival tech company Anthropic announced a limited release of its own highly capable cybersecurity tool named Mythos.
The new OpenAI model serves a very specific purpose. The company designed GPT-5.4-Cyber to help large organizations hunt for hidden software issues so they can fix them before hackers find them. To make the tool more effective, OpenAI removed several traditional safety constraints. The new model places fewer limits on how users can probe the system, allowing security professionals to test the artificial intelligence in much more creative and aggressive ways.
Not everyone can access this powerful new tool just yet. OpenAI will offer the model exclusively to participants of its Trusted Access for Cyber program. The company launched this specific security program in February to allow trusted corporate customers and professional cybersecurity researchers to test its most capable and dangerous software offerings in a controlled environment.
OpenAI plans to slowly increase the number of people who can use the new model. During the initial rollout phase, the company will let several hundred vetted users test the system and provide feedback. Over the coming weeks, OpenAI will expand its early access program to include thousands of approved cybersecurity professionals worldwide.
A fierce race is currently underway in the technology sector. Both OpenAI and Anthropic are spending billions of dollars to develop advanced artificial intelligence models capable of handling incredibly complex tasks. Both companies desperately want to convince massive global businesses to pay high subscription fees for these powerful tools. However, as these software models improve at writing code and detecting security flaws, government officials are sounding serious alarms.
Security experts worry deeply about how people might misuse this new technology. If these powerful coding tools fall into the wrong hands, state-backed hackers or organized criminal groups could use them to find and exploit vulnerabilities in banking systems or national infrastructure. Criminals already use basic artificial intelligence programs to help write malware and launch sophisticated cyberattacks against everyday people and large corporations.
Anthropic threw fuel on the fire just last week. The company introduced its own specialized artificial intelligence model called Mythos. Anthropic executives bragged that Mythos specializes specifically in identifying and exploiting deep vulnerabilities across massive operating systems and popular web browsers. Because the software is so dangerous, the company decided to limit the initial release heavily. Anthropic only shared the tool with a very select group of highly trusted corporate partners, including tech giants like Amazon, Apple, and Microsoft. The goal is to enable these massive companies to use Mythos to spot hidden issues and safeguard their internal networks.
However, the release of Mythos quickly sparked intense panic across the financial and government sectors. The US government takes the threat very seriously. Last week, US Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell summoned top Wall Street leaders for an emergency meeting. During the briefing, government officials explicitly warned the bank executives that they must take the new Mythos model seriously and prepare their digital defenses accordingly.
The government does not want to fall behind the private sector. The Treasury Department’s own internal technology team actively reached out to Anthropic. The government hackers want to gain full access to the Mythos software so they can begin hunting for critical vulnerabilities inside federal computer systems before foreign enemies find them first.
