OpenAI Receives Federal Approval to Launch GPT-5.6 Models Thursday

Key Takeaways

• OpenAI received federal authorization to deploy GPT-5.6 publicly this Thursday, featuring Sol, Terra, and Luna model variants
• Security concerns regarding AI’s capacity to detect software weaknesses caused the initial postponement
• Federal officials under the Trump administration authorized widespread deployment after evaluations and consultations with OpenAI leadership
• Competitor Anthropic experienced temporary suspension of its Mythos 5 and Fable 5 systems following June 12 export restrictions, with limited restoration occurring recently
• OpenAI and Anthropic have both submitted private IPO filings seeking approximately $1 trillion valuations

OpenAI has announced it will deploy its most advanced artificial intelligence model collection, GPT-5.6, this Thursday after receiving federal authorization following comprehensive national security evaluations.

The GPT-5.6 collection consists of three distinct tiers. Sol represents the premium flagship offering. Terra serves as the intermediate option designed for routine applications. Luna provides a rapid, economical alternative.

The company announced the deployment through a Tuesday post on X, noting it was simultaneously broadening preview availability across international markets before the complete rollout.

The deployment experienced postponement after federal authorities requested OpenAI restrict availability to a limited group of authorized domestic partners. This directive emerged from apprehensions that sophisticated AI systems might enable malicious actors to discover code vulnerabilities.

According to Axios, the Trump administration authorized the expanded deployment following supplementary technical evaluations and discussions between OpenAI representatives and government authorities.

Security Considerations Influenced Deployment Timeline

Washington has intensified monitoring of advanced AI systems due to concerns about potential exploitation by foreign adversaries, particularly China and Russia.

The United States and China are engaged in competitive development of next-generation AI capabilities, with analysts warning these systems could enable more sophisticated cyber operations targeting complex, outdated infrastructure networks.

OpenAI competitor Anthropic encountered comparable limitations. Its Mythos 5 and Fable 5 systems were deactivated for all users following a federal export control directive issued June 12. Limited access was reestablished recently after Anthropic implemented protective measures.

Mythos, designed specifically for cybersecurity specialists, continues to be accessible exclusively to approved US-based organizations. Chinese officials have expressed concerns the system could be weaponized to exploit software weaknesses against their strategic interests.

Anthropic has stated it is “probably impossible” to create any AI system completely immune to circumvention attempts.

Cost Structure and Market Dynamics

OpenAI announced Terra will carry pricing at fifty percent of its predecessor, GPT-5.5, as rivalry with Anthropic and Google escalates.

GPT-5.6 Sol was characterized by OpenAI as comparable to Anthropic’s Mythos Preview on the ExploitBench cybersecurity evaluation metric during late June previews.

Elon Musk’s SpaceXAI also took action this week, releasing its Grok 4.5 system to general users.

The Trump administration has enacted an executive directive establishing a voluntary arrangement allowing AI developers to provide frontier systems to federal authorities for up to 30 days prior to public deployment.

OpenAI has expressed opposition to making this type of government review process the permanent standard, contending it delays access to powerful capabilities for enterprises and consumers who require them.

Both OpenAI and Anthropic have submitted confidential IPO documentation with federal regulators, with each company pursuing valuations approaching $1 trillion.

The post OpenAI Receives Federal Approval to Launch GPT-5.6 Models Thursday appeared first on Blockonomi.