All the athletes are at the starting blocks: the race for the most advanced models is underway, while governments are concerned that they do not threaten national security. The new artificial intelligence models Advanced AI is at the heart of the AI boom, powering autonomous tasks like writing software code, generating content, and handling customer support requests, as Meta, Mask, Anthropic, and OpenAI are doing. Conversely, AI is intensifying. government protective actions on suspicion of threats from the national security, as is happening reciprocally in the USA and in China.
Meta Launches Muse Image and Soon Muse Video
Meta Platforms, the owner of Facebook and Whatsapp has announced the launch of Muse Image, the first image generation model that will be implemented in the future with Muse VideoThe new tool, developed by Meta Superintelligence Labs and integrated into its Meta AI chatbot, it is able to satisfy complex requests, use photos as input and allow users to edit images based on simple sketches or notes. Last April, the company had launched Muse Spark, the dedicated AI model instead to textual analysis and reasoning.
In particular, Muse Image will power over 30 new AI-powered effects for Instagram Stories and will enable the generation of images in direct chats with Meta AI on WhatsAppThis will initially only happen in some countries but the company intends to expand Muse Image to other countries and integrate it into Facebook and Messenger.basic usage of Muse Image with Meta AI will be Free, while some other creative features will be available through Meta's subscription plans, the company said.
OpenAI: Trump clears the powerful GPT Sol. Low-cost versions of Terra and Luna are also coming.
OpenAI will officially launch its most powerful Artificial Intelligence model yet tomorrow, the so-called GPT-5.6 Sol, after having postponed its launch last month at the request of the U.S. government, due to the growing national security concerns relating to the possible improper use. He reports it Axios adding that the Trump administration gave the green light after further testing and meetings between the company and government officials. In addition to its highest-performing model, GPT-5.6 Sol, Meta will also launch the cheaper Earth and Moon models, as announced in a post on X last night.
Allbeginning of the month OpenAI had proposed selling a 5% stake in the company to the U.S. government. The move, according to the Financial Times, It would aim to reduce the growing political pressure on Washington and open up a new model for sharing the economic benefits of artificial intelligence.
Musk responds with the launch of his “faster, more efficient and less expensive” Grok 4.5
Elon Musk is also on the launch pad. His SpaceX will make its artificial intelligence model available to the public tomorrow. Grok 4.5, as announced by the CEO in a social media post last night. Musk described the model as “Opus-class,” comparing it to the Claude from anthropic, but claimed it is “faster, more token-efficient, and less expensive.” Grok is the flagship AI model of XAI, Musk's company that was merged into SpaceX this year and recently renamed SpaceXAI. xAI released its latest version, Grok 4.3, in April, with Musk having long anticipated the next iteration of his flagship AI model.
But Musk's aces up his sleeve aren't over. According to what was reported by The Information, which cites a note sent to staff, SpaceXAI and Cursor They plan to launch their first jointly developed AI model soon, and the new model is expected to process information quickly, making it competitive in some respects with Opus 4.8. anthropic and GPT 5.5 of OpenAIIn June, SpaceX announced the acquisition of Anysphere, the startup that created the popular AI-based programming agent. Cursor, in a $60 billion, all-stock-financed deal.
National security concerns in the US as well as in China
Meanwhile the United States and China are engaged in a race for development of cutting-edge artificial intelligence models also for reasons of mutual national security. Washington has stepped up its control on versions of advanced artificial intelligence models to identify potential threats, fearing that the technology could be misused by militar o intelligence services in China, Russia or other countries.
Trump just signed a executive order which establishes a picture under which AI developers can provide the U.S. government with “covered frontier models” for up to 30 days before releasing them to other trusted partners. anthropic, a competitor of OpenAI, had suddenly disabled its most advanced AI models, Mythos 5 e Fable 5, for all users after theUS government order on June 12, regarding export controls, citing national security concerns. The restrictions were lifted last week after Anthropic implemented certain security measures.
On the other side of the Pacific, the Chinese authorities are concerned about the possibility that the same Mythos exploits software vulnerabilities and that the United States could use this model against Chinese interests. Chinese authorities have held meetings with major technology companies regarding the possibility of restrict access from abroad to China's most advanced artificial intelligence models, including those not yet made public.
This morning, the cybersecurity platform run by the Chinese Ministry of Industry warned that it had identified a serious risk of “backdoor” in the of your digital ecosystem. of the AI programming tool anthropic, known by the name of Claude Code. In a statement posted on his account WeChat, the National Vulnerability Database said that Claude Code contains a built-in monitoring mechanism capable of transmitting sensitive information, including the position geographic location of users and identifiers relating to theidentity, to remote servers without users' consent.
Nvdb advised organizations and users to immediately examine affected systems and uninstall the incriminating versions or update to the latest secure version in which the alleged backdoor code has been removed. The Chinese giant Alibaba has banned employees from using Claude Code at work after the tool came under scrutiny for features that can help identify users linked to China, as reported by Reuters last week.
