Trump puts brakes on OpenAI’s newest AI model

OpenAI's newest AI exemplary is getting a gatekeeper earlier astir radical tin effort it: the U.S. government.

The institution says GPT-5.6 Sol, its newest flagship model, volition commencement with a constricted preview for a tiny radical of trusted partners. OpenAI said it previewed the model's capabilities with the U.S. authorities up of motorboat and, astatine the government's request, is starting with a constricted radical of trusted partners whose information has been shared with the authorities earlier a broader rollout. The exemplary volition reportedly beryllium accessible lone to customers approved by the Trump medication during this archetypal preview period.

MOST PROMINENT AI CHATBOTS HAVE LIBERAL BIAS, NEW STUDY FINDS

That to maine is simply a large shift. Usually, you perceive astir a caller AI exemplary and wonderment erstwhile you tin usage it. Now, the bigger question whitethorn beryllium who gets entree archetypal and however those decisions are made.

Sign up for my FREE CyberGuy Report

• Get my champion tech tips, urgent information alerts and exclusive deals delivered consecutive to your inbox.
• For simple, real-world ways to spot scams aboriginal and enactment protected, sojourn CyberGuy.com - trusted by millions who ticker CyberGuy connected TV daily.
• Plus, you'll get instant entree to my Ultimate Scam Survival Guide escaped erstwhile you join.

OpenAI is limiting entree to GPT-5.6 Sol arsenic U.S. officials reappraisal imaginable cybersecurity risks tied to the almighty caller AI model. (Kurt "CyberGuy" Knutsson)

What is OpenAI's GPT-5.6 Sol?

GPT-5.6 Sol is portion of OpenAI's caller GPT-5.6 lineup, which besides includes Terra and Luna. OpenAI describes Sol arsenic its strongest exemplary yet, with improvements successful coding, biology and cybersecurity tasks. In its nationalist announcement, OpenAI said GPT-5.6 models volition initially beryllium disposable done the API and Codex to a prime radical of trusted partners and organizations, with broader availability planned for ChatGPT, Codex and the API soon. As of aboriginal July, OpenAI says GPT-5.6 isn’t disposable successful ChatGPT during the preview, and the institution has not announced a general-availability date.

The cybersecurity portion is what has Washington paying adjacent attention. OpenAI says Sol is amended astatine helping radical find and hole bundle vulnerabilities than reliably carrying retired afloat cyberattacks. The institution besides says Sol does not transverse its interior "Cyber Critical" threshold, though it admits benchmark tests cannot foretell each imaginable usage erstwhile the exemplary is combined with different tools.

That is the concern. The aforesaid AI that tin assistance a information squad hole a bundle flaw could besides springiness scammers and hackers a caput commencement if the guardrails fail.

Why the authorities is stepping in

The Trump medication has been gathering a caller AI cybersecurity framework. A June enforcement bid says precocious AI tin fortify the country, but it besides creates nationalist information concerns that necessitate coordination crossed national agencies and backstage companies.

The White House information expanse says the bid creates a voluntary model for covered frontier models. It besides says the authorities whitethorn get unafraid aboriginal entree for trusted partners, portion stating the bid does not authorize mandatory licensing, pre-clearance oregon permits for releasing AI models.

That favoritism is important. The medication says this is astir security. Critics interest it could crook into quiescent authorities power implicit who gets the newest AI tools first.

Anthropic was pulled into the aforesaid AI fight

OpenAI isn’t the lone AI institution that has faced authorities unit implicit almighty AI models. On June 12, Anthropic said the U.S. authorities directed it to suspend entree to its Fable 5 and Mythos 5 models by overseas nationals. Anthropic said the bid forced it to disable the models much broadly truthful it could comply.

CyberGuy antecedently covered Anthropic's Mythos AIand wherefore its quality to find bundle vulnerabilities raised large cybersecurity questions. That helps explicate wherefore these models became portion of the larger statement implicit AI, cyber hazard and who gets entree first.

Since then, the standoff has changed. The U.S. Commerce Department aboriginal lifted export controls connected Anthropic's Fable 5 and Mythos 5 models. Anthropic said it would reconstruct access, and Fable 5 was reportedly backmost online arsenic of July 1. Reports accidental Anthropic agreed to fortify safeguards, enactment with the U.S. authorities connected model-release protocols and study malicious enactment tied to the models. Reuters besides reported that the restrictions were removed aft talks betwixt Anthropic and U.S. officials.

That makes the Anthropic occurrence a utile examination for OpenAI's constricted rollout. In some cases, the contented is nary longer lone astir what an AI exemplary tin do. It is besides astir however rapidly almighty AI tools should beryllium released, who should get aboriginal entree and what relation the authorities should play.

OPENAI'S SAM ALTMAN WANTS TO NEGOTIATE A 5% STAKE IN COMPANY FOR US IF COMPETITORS AGREE TO KEY PROVISION

The Trump administration’s AI cybersecurity propulsion is raising caller questions astir who gets aboriginal entree to precocious artificial intelligence. (Samuel Boivin/NurPhoto via Getty Images)

Why this reaches beyond Silicon Valley

Here is the bigger question: however should almighty AI tools beryllium released erstwhile they tin assistance support hospitals, banks and businesses, but could besides assistance the incorrect idiosyncratic find information holes faster?

Security teams privation amended tools due to the fact that they are already stretched thin. Hospitals and wellness attraction systems,banks, section utilities and captious systems, positive businesses each tally connected bundle that tin person hidden holes. A stronger AI exemplary could assistance find those anemic spots earlier criminals exploit them.

However, the entree question is complicated. Who gets aboriginal access? Who gets told to wait? What happens if a tiny information institution needs the aforesaid AI instrumentality arsenic a large authorities contractor?

For now, the statement comes down to balancing security, entree and accountability arsenic these AI tools determination person to the services you trust connected each day.

How to support yourself arsenic AI gets much powerful

AI tools are getting much capable, and that means scammers volition support utilizing the buzz astir them to instrumentality you. Here are the steps to instrumentality now.

1) Watch retired for fake early-access offers

Whenever a large AI merchandise gets attention, imposters follow. Be cautious with emails, texts oregon ads that committedness aboriginal entree to GPT-5.6 Sol, Mythos 5 oregon immoderate restricted AI model. Go straight to the authoritative institution tract alternatively of clicking a link.

2) Use beardown antivirus software

Strong antivirus bundle tin assistance support you from malicious links, phishing sites and downloads that scammers whitethorn disguise arsenic AI tools oregon beta invitations. The champion mode to support yourself from malicious links that instal malware, perchance accessing your backstage information, is to person beardown antivirus bundle installed connected each your devices. Get my picks for the champion 2026 antivirus extortion winners for your Windows, Mac, Android & iOS devices astatine Cyberguy.com

3) Use a password manager

Use strong, unsocial passwords for your important accounts. A password manager tin assistance you make and store analyzable passwords, truthful you are not reusing the aforesaid login crossed aggregate sites.

4) Turn connected two-factor authentication

Turn connected two-factor authentication (2FA) whenever it is available. Authentication apps oregon information keys are stronger than substance connection codes due to the fact that they are harder for scammers to intercept.

OPENAI TIGHTENS AI RULES FOR TEENS BUT CONCERNS REMAIN

OpenAI says GPT-5.6 Sol tin assistance find and hole bundle flaws, but officials interest almighty AI tools could besides assistance hackers. (Photo by Philip Dulian/picture confederation via Getty Images)

5) Keep your devices and apps updated

Keep your phone, computer, browser and apps updated. Many cyberattacks people known bundle flaws that already person patches, truthful updates are 1 of the easiest ways to adjacent information gaps.

6) Be cautious what you upload into AI tools

Think doubly earlier pasting delicate accusation into immoderate AI tool. Avoid uploading Social Security numbers, fiscal records, aesculapian documents oregon backstage institution files unless you cognize precisely however the work handles your data.

7) Remove your idiosyncratic information from people-search sites

Scammers tin marque AI-themed attacks much convincing erstwhile they tin find your location address, telephone number, relatives oregon workplace online. A information removal work tin assistance trim however overmuch of your idiosyncratic accusation appears connected information broker and people-search sites. Check retired my apical picks for information removal services and get a escaped scan to find retired if your idiosyncratic accusation is already retired connected the web by visiting Cyberguy.com

8) Slow down earlier clicking urgent AI invites

Scammers emotion urgency. If an connection claims you request to enactment accelerated to get entree to a caller AI model, dainty that arsenic a informing sign. Search for the announcement yourself and verify it from the company's authoritative website.

Watch the CyberGuy Live replay: Lock Down Your Phone successful 30 Minutes

Your telephone holds your email, passwords, photos, banking apps and idiosyncratic data. In this escaped CyberGuy Live replay, Kurt the CyberGuy walks you measurement by measurement done elemental telephone information fixes you tin bash astatine your ain pace. You’ll larn however to amended your privateness settings, spot the latest telephone scams, usage trusted information tools and locomotion distant with a elemental checklist to enactment protected. Watch the replay and get our checklist here: CyberGuyLive.com

Kurt's cardinal takeaways

AI has reached a constituent wherever Washington wants a person look earlier everyone gets access. I recognize wherefore cybersecurity officials are paying attention. A exemplary that tin spot bundle flaws could assistance support a hospital, a slope oregon a utility. In the incorrect hands, though, that aforesaid capableness could make caller risks. For you, the takeaway is straightforward. AI tools volition support getting much capable, and that tin beryllium bully quality erstwhile they assistance support your data. However, you inactive request to support your accounts, question astonishment AI offers and deliberation doubly earlier sharing delicate accusation with immoderate caller tool.

Should almighty AI tools look a information reappraisal earlier nationalist release, oregon should companies determine erstwhile they are acceptable for everyone to use? Let america cognize by penning to america astatine Cyberguy.com

CLICK HERE TO DOWNLOAD THE FOX NEWS APP

Sign up for my FREE CyberGuy Report

• Get my champion tech tips, urgent information alerts and exclusive deals delivered consecutive to your inbox.
• For simple, real-world ways to spot scams aboriginal and enactment protected, sojourn CyberGuy.com - trusted by millions who ticker CyberGuy connected TV daily.
• Plus, you'll get instant entree to my Ultimate Scam Survival Guide escaped erstwhile you join.

Copyright 2026 CyberGuy.com. All rights reserved.