AI news · Wednesday, July 1, 2026

Anthropic regains export rights after adding security safeguards to AI models

The US government has officially lifted export restrictions on Anthropic’s Mythos and Fable models, effectively allowing the company to resume international access after a standoff regarding national security risks. The breakthrough came after Anthropic agreed to deploy a new guardrail that forces user queries to be rerouted to a less-advanced model if they trigger specific security concerns, a move that satisfied Commerce Secretary Howard Lutnick’s team. This resolution highlights the precarious position of AI labs, which are increasingly caught between the rapid pace of their own innovation and an erratic regulatory environment.

Meanwhile, a different kind of tech vulnerability made headlines when security researcher Ian Carroll used Claude to identify a flaw in Front Gate Tickets, a subsidiary of Live Nation. By exploiting a nested SQL query—a technical way of tricking a database into running unintended commands—the AI helped Carroll bypass firewall protections to access sensitive staff data and generate nearly unlimited free VIP tickets for major US festivals. The vulnerability was resolved within 24 hours, but the incident underscores growing fears that AI can act as a force multiplier for digital security threats.

In the corporate world, the AI gold rush continues to drive significant capital reshuffling. Bending Spoons, an Italian firm that buys and revitalizes aging software brands like Evernote and Vimeo, had a successful IPO, closing 40% above its offering price as it leans into an acquisition-heavy strategy fueled by AI efficiency. Elsewhere, Meta is reportedly eyeing its own cloud infrastructure business, aiming to monetize its massive surplus of AI compute power by renting it out to others, much like SpaceX has done with its own data centers.

This trend of turning infrastructure into a cash cow arrives as some market analysts, including those at JPMorgan, warn that the massive divergence between rising chip stocks and struggling AI-spending firms echoes the dangerous market dynamics of the dot-com era. While giants like Oracle and Intel continue to pour billions into infrastructure and AI development, they are simultaneously executing deep workforce cuts to rebalance their budgets. As these companies pivot toward leaner operations, the debate over the sustainability of this spending continues to intensify, with some industry figures like Palantir CEO Alex Karp arguing that many enterprises are paying for AI 'tokens'—the building blocks of model output—that simply provide no measurable value to their bottom line.

The quick hits

Anthropic is restoring international access to its Mythos and Fable models after reaching a security agreement with the US government — ending a weeks-long regulatory freeze.
A security researcher used Claude to find a flaw in a major ticket vendor that allowed for the creation of unauthorized VIP concert passes — showing how AI can help identify and exploit software bugs.
Bending Spoons went public at an $18 billion valuation — signaling that investors are still betting on companies that use AI to aggressively cut costs at older software brands.
Meta is exploring a new cloud business to rent out its excess AI computing power — mirroring a strategy being used by SpaceX to recoup some of its massive infrastructure investments.
Major tech firms including Intel and Oracle are spending billions on layoffs as they pivot to AI — illustrating the human cost of the industry's massive reallocation of resources.

Anthropic regains export rights after adding security safeguards to AI models

The quick hits

Sources