Anthropic Mythos Release, OpenAI Restrictions, Meta Smart Glasses — AI News Briefing

Top 7 Stories

1. US Government Approves Anthropic’s “Mythos” AI Model for Trusted Organizations

The U.S. government has authorized Anthropic to release its powerful new frontier model “Mythos” to a select group of trusted U.S. companies and government agencies. Reported by Reuters, CNBC, and Semafor, the decision marks a significant milestone in the government’s evolving approach to frontier AI oversight — opting for controlled, trusted access rather than outright restriction.

Mythos is understood to be Anthropic’s most capable model to date, and the release comes with strict usage guardrails. The arrangement effectively creates a government-vetted tier of AI access, raising questions about how “trusted” status is determined and whether this model sets a precedent for future frontier model releases. Organizations granted access will be subject to ongoing monitoring and auditing requirements.

2. US Tells OpenAI to Restrict Access to Its Most Powerful AI Model

In a parallel move, the U.S. government has directed OpenAI to restrict access to its most advanced AI model, according to a Computerworld report from June 26. While details remain sparse, the directive appears grounded in national security concerns about the model’s capabilities in areas like autonomous cyber operations and bioweapon design.

This creates an interesting policy juxtaposition: while Anthropic’s Mythos gets a controlled green light, OpenAI faces a more restrictive posture. The differential treatment may reflect varying levels of safety infrastructure, model architecture choices, or government confidence in each company’s alignment practices. OpenAI has not publicly commented on the specifics of the restriction.

3. Anthropic Accuses Alibaba of Massive Claude AI Data Scraping Operation

Anthropic has publicly accused Chinese tech giant Alibaba of orchestrating a “brazen” campaign to steal capabilities from its Claude AI model, using an estimated 25,000 fake accounts to systematically scrape outputs. The Wall Street Journal, AI Business, and InfoWorld report that Anthropic detected the operation and has taken countermeasures.

The incident underscores the growing challenge of model theft and intellectual property protection in the AI industry. Alibaba has not yet issued a detailed response, but the accusation comes amid escalating U.S.-China tech tensions and could have significant implications for cross-border AI collaboration and data security practices going forward.

4. OpenAI Leans Toward Holding Up IPO Until Next Year

The New York Times reports that OpenAI is leaning toward delaying its highly anticipated initial public offering until 2027. The decision, attributed to sources familiar with internal discussions, suggests the company wants more time to mature its revenue streams and navigate the increasingly complex regulatory landscape.

An OpenAI IPO has been one of the most closely watched events in tech finance, with valuations speculated to rival major public companies. A delay could signal that the company is prioritizing stability and long-term positioning over near-term capital raises, or that it anticipates significant regulatory headwinds that would complicate a 2026 public listing.

5. Meta Launches $299 AI Smart Glasses with “Muse Spark” Assistant

Meta has officially launched its self-branded AI smart glasses starting at $299, featuring the new Muse Spark AI assistant. The glasses integrate directly with Meta’s AI ecosystem, offering real-time visual recognition, natural language interaction, and contextual AI assistance without requiring a smartphone tether for core functionality.

The launch marks Meta’s most aggressive push into AI wearable hardware yet. At $299, the price point undercuts competitors significantly, positioning the glasses as a mass-market consumer device rather than a niche developer tool. Early reviews highlight the Muse Spark assistant’s natural conversation flow and the glasses’ all-day battery life as key differentiators.

6. Google Delays Gemini 3.5 Pro Launch to July

Google has pushed back the launch of its next-generation frontier model, Gemini 3.5 Pro, to July as the company continues fine-tuning its capabilities, according to Business Insider. The delay comes as Google aims to ensure the model meets its safety and performance benchmarks before public release.

The extra development time signals that Google is being cautious about quality and safety after the mixed reception of some earlier Gemini iterations. The delay also means Google’s next major model release will arrive amid an increasingly crowded field, with OpenAI, Anthropic, and others all advancing their own frontier capabilities.

7. Five Eyes Intelligence Alliance Warns AI Models Capable of “Devastating Attacks” Months Away

In a rare joint statement, the Five Eyes intelligence alliance (U.S., UK, Canada, Australia, New Zealand) has warned that AI models capable of launching devastating autonomous cyberattacks on governments and businesses could be available within months. The Guardian reports the warning highlights the accelerating pace of AI capabilities outpacing defensive measures.

The statement calls for urgent international coordination on AI safety standards and monitoring. Security experts note that the public nature of the warning is itself significant — intelligence alliances typically avoid disclosing capability timelines, suggesting the perceived threat level is unusually high. The warning is expected to intensify calls for AI regulation and frontier model oversight.

Trend Watch

What to Watch

The government’s dual-track AI regulation strategy is the story to watch this week. The contrasting approaches to Anthropic (controlled release) and OpenAI (restriction) will likely define how frontier AI models are managed going forward. Industry observers should watch for formal policy frameworks emerging from this ad-hoc approach.

AI hardware race heats up. Meta’s $299 smart glasses launch, combined with ongoing developments from Apple and others, suggests 2026 could be the year AI wearables truly enter the consumer mainstream. Watch for developer ecosystem announcements and third-party app integrations.

U.S.-China AI IP tensions. The Anthropic-Alibaba scraping allegations add a new dimension to the tech rivalry. Expect more formal mechanisms for model protection, potential export control expansions, and increased scrutiny of cross-border AI data flows.

Frontier model release timelines. With Gemini 3.5 Pro delayed to July and Mythos rolling out under controlled access, the next 30 days will be revealing for how the major labs sequence their most powerful releases amid evolving government oversight.