Amazon CEO Andy Jassy may have been the source of security concerns that led Anthropic to cut off worldwide access to two models on Friday.

The Wall Street Journal reports that Jassy told Treasury Secretary Scott Bessent and other government officials that Amazon researchers used Anthropic’s Claude Fable 5 to obtain information that could be used in cyberattacks. The government subsequently imposed an export control ban on the Fable 5 and Mythos 5 models.

An Amazon spokesperson said in a statement that while it’s “not uncommon for governments to seek our counsel on potential security risks,” the company does not “share the details of those discussions.”

The spokesperson also pointed to an update stating that AWS has been affected by the model cut off.

The Information and Reuters similarly reported that Amazon (a major Anthropic investor) had communicated concerns about the security of Anthropic’s models.

David Sacks, Trump’s former AI czar who now co-chairs the President’s Council of Advisors on Science and Technology, offered his own account of the discussions, claiming that “a highly credible trusted partner of both Anthropic and the USG […] came forward with a jailbreak.”

Sacks added, “The Admin asked [Anthropic CEO Dario Amodei] to fix the jailbreak or de-deploy the model. Dario refused.”

Anthropic said in a blog post that the capabilities apparently causing government concern are already available in other publicly accessible models.

This post has been updated with a statement from an Amazon spokesperson.

Meta has begun dismantling its $2 billion acquisition of Manus, completing an operational separation from the Chinese-founded AI startup and halting data sharing between the two companies. This is the most concrete step yet toward complying with a divestiture order Beijing issued roughly two months ago on national security grounds.

Meta has cut Manus off from its internal systems, Bloomberg reported, preventing employees from using Manus tools for internal projects as the two companies move toward a full separation.

Meanwhile, according to May reports, the co-founders of Manus have held preliminary discussions about raising approximately $1 billion from outside investors to reclaim the startup from Meta, a move that could pave the way for a Chinese joint venture structure and an eventual listing in Hong Kong, a venue that has seen a surge in AI listings this year for Chinese AI startups like MiniMax and Zhipu.

What was supposed to be a landmark exit for Chinese AI is quickly unraveling. The move underscores Beijing’s determination to retain control over strategically sensitive technology, regardless of a company’s offshore incorporation.

In addition to the forced divestiture, Chinese authorities have since expanded travel restrictions to researchers and executives at private firms, requiring government approval before heading abroad. China is also tightening its grip on foreign capital, with reports indicating that top AI firms, including Moonshot AI, StepFun, and ByteDance, will need government sign-off before accepting U.S. investment, adding another layer to Beijing’s sweeping effort to control its AI sector.

Even as Meta moves to sever ties with Manus, the agentic AI startup has continued to ship new features, rolling out integrations with Similarweb and Shopify.

Manus drew widespread attention with a viral agent demo relocated its staff to Singapore in mid-2025 before announcing a $2 billion acquisition by Meta in December. Chinese regulators moved to scrutinize the transaction earlier this year, citing potential violations of technology export controls and foreign investment rules.

Manus investors, including California-based venture firm Benchmark, have already received their proceeds from the acquisition, while Asian backers, including Tencent, HSG, and ZhenFund, have indicated they will cooperate with the unwinding process, according to the WSJ.

Manus’ Chinese origins with parent company Butterfly Effect drew scrutiny on both sides of the Pacific, with Senator John Cornyn questioning whether American capital should flow to a Chinese-linked firm.

Meta and Manus did not immediately respond to a request for comment outside regular business hours.

Astronauts aboard the International Space Station reportedly took precautionary shelter measures after maintenance work on a long-troubled Russian module raised fresh safety concerns about air leaks aboard the orbiting laboratory.

According to reports, the incident involved Russia’s Zvezda service module, which has experienced recurring air leak problems for several years. During repair work and pressure tests connected to the leak investigation, astronauts were instructed to isolate themselves in safer sections of the ISS as engineers monitored the station’s integrity and pressure stability.

The event highlights the growing strain on aging infrastructure aboard the International Space Station, particularly within some of the older Russian-built modules that have remained operational far beyond their original expected service life. While NASA and Roscosmos reportedly emphasized that the crew was never in immediate danger, the precautionary shelter procedures underscore how seriously space agencies continue to treat even relatively small leaks in orbit.

The International Space Station is showing signs of age

The Zvezda module has become a recurring source of concern in recent years. First launched in 2000, the module serves as one of the core components of the Russian segment of the ISS and supports life-support systems, crew quarters, and station operations.

Air leaks linked to tiny cracks inside the module have been detected multiple times over the past few years. While engineers have repeatedly patched and monitored the affected areas, the exact cause of some leaks remains difficult to fully determine. Reports suggest that Russian teams recently conducted additional sealing efforts and pressure tests to determine whether the repairs had successfully stabilized the module.

Image: The long axis of the complex is seen in this 2021 image of the International Space Station. In this picture, the US segment and Crew Dragon docking port are close to the top, while the Russian segment is toward the bottom. NASA is credited.

Because the ISS operates in the vacuum of space, even minor air leaks are treated as serious engineering risks. A gradual loss of cabin pressure can threaten crew safety if not carefully monitored and controlled. Astronauts regularly train for emergency leak scenarios, including isolating modules and taking shelter in docked spacecraft if necessary.

The Zvezda service module transfer tunnel, known as PrK, has suffered from cracks and leaks for some time, and has been mitigated by Roscosmos as much as possible to date. The cracks have always been a concern that NASA watches very closely. NASA and Roscosmos have been working…

— Bethany Stevens (@NASASpox) June 5, 2026

The incident also occurred during a period of increasing uncertainty surrounding the future of the ISS itself. Originally designed with a more limited operational lifespan, the station has now been continuously inhabited for more than two decades. As components age, maintenance challenges are becoming more frequent and more technically complex.

Space agencies are trying to keep the ISS running safely

Despite the concerns, the ISS continues to function as one of humanity’s most important scientific research platforms. NASA, Roscosmos, ESA, JAXA, and other international partners still rely heavily on the station for microgravity experiments, medical research, and long-duration spaceflight studies.

However, recurring technical issues have intensified discussions about how long the station can realistically remain operational before replacement systems become necessary. NASA has already shifted focus toward supporting future commercial space stations that could eventually succeed the ISS later in the decade.

For now, engineers appear focused on stabilizing the Zvezda module and continuing close monitoring of pressure levels and structural integrity. Reports indicate the station remains operational, and astronauts have resumed normal activities following the precautionary shelter procedures.

Still, the episode serves as another reminder that maintaining a permanently occupied laboratory in space is an extraordinarily difficult task – especially when parts of that laboratory are now older than many of the astronauts living inside it.

What started as another Minecraft modding campaign has turned into one of the more disturbing malware stories tied to gaming communities this year. Security researchers at McAfee have uncovered a large-scale malware operation called “WeedHack” that reportedly infected more than 116,000 devices, primarily targeting Minecraft players through fake mods, cheats, cracked software, and community tools.

But unlike many traditional malware campaigns focused purely on stealing passwords or crypto wallets, this one appears to have crossed into harassment, cyberbullying, and invasive surveillance.

According to McAfee’s research, the campaign was allegedly operated by a teenager and relied heavily on Discord, Minecraft communities, and social engineering tactics to spread malicious files disguised as popular game-related downloads.

A malware campaign built around Minecraft communities

Researchers say WeedHack spread through malicious Minecraft mods, cheats, installers, macros, and pirated tools shared across Discord servers and gaming forums. Victims believed they were downloading performance mods or gameplay enhancements, but instead installed malware capable of stealing personal information and remotely accessing systems.

The malware reportedly harvested browser credentials, Discord tokens, crypto wallet information, screenshots, and personal files. In some cases, infected users were allegedly blackmailed, harassed, or publicly humiliated using stolen private information.

McAfee says the operation relied heavily on young gamers trusting files shared within online communities. Minecraft remains one of the world’s largest gaming ecosystems, with millions of active players and an enormous modding culture that frequently involves downloading third-party software from unofficial sources.

That openness created an ideal environment for the malware campaign to spread rapidly.

The report also highlights how modern malware operations are increasingly blending cybercrime with online harassment culture. Researchers say some victims experienced targeted bullying and intimidation after infection, making the campaign more invasive than typical, financially motivated attacks.

The malware reportedly evolved constantly to avoid antivirus detection, with operators updating payloads and distribution methods across multiple platforms.

Why this campaign is particularly concerning

Gaming communities have increasingly become major targets for cybercriminals because younger users often install unofficial files more casually than enterprise users or experienced professionals.

Minecraft, in particular, has one of the largest user-generated content ecosystems in gaming, making it extremely difficult for players to distinguish safe mods from malicious downloads.

The scale of the WeedHack campaign also shows how accessible cybercrime tools have become. Researchers suggest the operation did not require sophisticated state-backed infrastructure or advanced hacking resources to infect over 116,000 devices globally.

More importantly, the campaign highlights how malware is evolving beyond simple financial theft. Cybercriminals are increasingly weaponizing personal information, Discord access, screenshots, and online identities for harassment and social manipulation.

What players should do

McAfee recommends Minecraft players avoid downloading mods, cheats, or cracked software from unofficial Discord servers or unknown sources. Users are also advised to enable multi-factor authentication, regularly scan devices for malware, and avoid reusing passwords across gaming platforms.

The report also serves as a reminder that gaming platforms are no longer isolated from broader cybersecurity threats. Online gaming communities now function much like social networks – and increasingly face the same risks tied to scams, surveillance, account theft, and coordinated abuse.

For Minecraft players, the bigger lesson is uncomfortable but increasingly important: the biggest danger online may no longer be the creepers inside the game, but the files being downloaded outside of it.

SoftBank, it seems, is not done buying robots. The Japanese group is in early talks to back a new funding round of around $800m (€700m) for Agile Robots, the Munich startup that builds robot arms, warehouse machines, and humanoids, according to Bloomberg.

SoftBank would contribute more than $300m of that total, people familiar with the discussions told the outlet, though they cautioned that the talks are at an early stage and the amounts and terms could still change.

Neither company would comment, and the round has not been independently confirmed beyond Bloomberg’s reporting, so the figures are best read as a snapshot of a negotiation rather than a done deal. What is on firmer ground is the relationship behind it.

SoftBank led Agile Robots’ $220m Series C in 2021, the round that turned the company into Germany’s first robotics unicorn, so a fresh cheque would be a return to a name already in the portfolio rather than a new acquaintance.

Agile Robots was formed in 2018 by researchers from the German Aerospace Center, the country’s space agency, and it now employs more than 3,200 people across Germany, China, and India.

It makes both the software and the hardware, which places it in the part of the robotics market investors have grown most excited about: machines that do physical work in factories, warehouses, and logistics depots, rather than chatbots that live on a screen.

That enthusiasm has money behind it. Global investment in robotics and physical AI reached $27.6bn in 2025, according to PitchBook, more than double the previous year’s total, as investors looked for ways to put AI to work in the physical world. The thesis, roughly, is that the software has matured and the next frontier is giving it hands.

For SoftBank’s founder, Masayoshi Son, the interest is not new. The company unveiled its Pepper humanoid back in 2014, then quietly halted production in 2020. More recently it has gone on the offensive, agreeing last year to buy Swiss group ABB’s industrial robotics arm for $5.4bn, and weighing a plan to create and list an AI and robotics company called Roze.

An Agile Robots cheque would slot neatly into that pattern, part of Europe’s quietly building robotics scene that Son keeps returning to, alongside his €75bn plan for AI data centres in France.

Agile Robots would also be joining a crowded European field. Germany’s Neura Robotics recently raised about €1bn, backed by stablecoin issuer Tether, at a roughly €4bn valuation, and Norwegian-American startup 1X was reported to be in talks for as much as $1bn at a $10bn valuation.

The money is arriving faster than the robots, which remains the recurring feature of the sector. For now, Agile Robots’ round is a conversation, with the cheques still being written.