America
US lifts export controls on Anthropic’s Claude 5 models after security upgrade
Anthropic has announced that the US government has lifted export controls on its newest artificial intelligence models, Claude Fable 5 and Claude Mythos 5, paving the way for the high-profile systems to return to active use.
The AI research and development company had temporarily suspended access for all users following the June 12 imposition of the restrictions, which blocked foreign nationals both inside and outside the US from accessing the models. Anthropic stated that the complete suspension was necessary because it was unable to perform real-time nationality verification. With the lifting of the export controls, the access restrictions have now been terminated.
Anthropic announced that Claude Fable 5 will be made available for global general availability starting Wednesday, July 1, across the Claude Platform, Claude.ai, Claude Code, and Claude Cowork. Users on Pro, Max, Team, and select Enterprise plans will be permitted to use Fable 5 for up to 50% of their weekly usage limits at no additional cost until July 7.
After that date, access will be transitioned to a usage-credit basis. The company is also working to restore access via Amazon Web Services (AWS), Google Cloud, and Microsoft Foundry as quickly as possible.
Meanwhile, access to Claude Mythos 5—a model designed for defense-oriented cybersecurity work that operates with fewer safety constraints—was restored to select US entities following approval from the US government on June 26.
The company stated that it continues to work in coordination with the US government to expand access to domestic and international partners under its Glasswing program.
The cybersecurity report that triggered export restrictions
According to the timeline disclosed by the company, the export controls were instituted on June 12, just three days after the models’ June 9 launch. The regulatory intervention was triggered by a report from Amazon researchers who successfully bypassed the safety guardrails of the Fable 5 model to identify specific software vulnerabilities.
The report highlighted a single instance where the model generated code demonstrating how the vulnerability in question could be exploited. Anthropic collaborated with the US government and partners, including Amazon, over a two-week period. During subsequent testing, Anthropic confirmed that lower-capacity models—such as Claude Opus 4.8, GPT-5.5, and Kimi K2.7—were also capable of identifying the same vulnerabilities.
Furthermore, the tests revealed that the code demonstrating how to exploit the vulnerability could be generated by all tested models, including Claude Haiku 4.5, Sonnet 4.6, Opus 4.7, GPT-5.4, and Kimi K2.7.
Anthropic maintained that the bypass technique did not expose any unique offensive cyber capabilities at the Mythos level, describing it instead as a borderline case involving routine defensive activities.
Nevertheless, in cooperation with the US government, the company trained an advanced safety classifier targeting the behavior identified in the report, which it claims blocks the exploit with a success rate of over 99%.
When cybersecurity-related queries directed at Fable 5 are blocked by this system, users will be notified and their prompts will be automatically redirected to the Opus 4.8 model.
Anthropic acknowledged that the new safety mechanism carries the risk of triggering more frequent false positives—blocking harmless queries during routine coding and debugging—but said it is optimizing the system to reduce these occurrences.
The company noted that Claude Mythos 5 is more capable of finding and exploiting cybersecurity vulnerabilities than any actor other than high-level human experts, making it an attractive target for malicious actors.
In contrast, Claude Fable 5 does not possess these offensive cyber capabilities. Anthropic stated that it doubled its staffing resources in the month leading up to the launch to implement its most robust cybersecurity measures to date on the model.
This approach, termed “defense-in-depth,” combines training the model to reject dangerous requests with the deployment of automated cybersecurity classifiers.
These classifiers, which function as smaller auxiliary AI systems, detect and block suspicious requests during cybersecurity operations. To counter the risk of “jailbreaking”—where users manipulate the system to bypass safety rules—Anthropic expanded its safety margins significantly.
While the company acknowledged that this conservative posture leads to the blocking of some harmless queries, it stated that the trade-off was necessary to prevent offensive cyber exploits.
Researchers at the Center for AI Safety and Innovation (CAISI), an agency under the US Department of Commerce, also evaluated the new measures and confirmed that the cybersecurity guardrails are highly robust.
A call for industry-wide safety standards
Pointing to a lack of objective, industry-wide standards to evaluate the severity of safety guardrail bypasses, Anthropic announced that it has begun developing a consensus-based framework alongside Amazon, Microsoft, Google, and other Glasswing partners.
The proposed cybersecurity assessment framework aims to score the severity of vulnerabilities based on four distinct criteria:
- Capability gain: The degree to which a vulnerability empowers a user beyond currently available tools.
- Scope of capability gain: The number of different offensive tasks that can be executed using the same bypass technique.
- Ease of weaponization: The level of human effort required to convert the method into an active exploit.
- Discoverability: How easily the technique can be identified and acquired.
Additionally, the company is launching a new HackerOne bounty program where security researchers can report cybersecurity vulnerabilities they discover within the Fable 5 model.
Deepening cooperation with the US government
Over the past ten weeks, Anthropic has worked closely with cybersecurity and national security agencies under the framework of the June 2 US Executive Order on “Promoting Advanced AI Innovation and Security.” The company announced it will further deepen its public-sector collaboration.
Under this expanded partnership, Anthropic will provide early, pre-launch access to critical national security-relevant models to government partners for independent testing.
The company has pledged to share information rapidly if vulnerabilities or abuse patterns are detected, and to submit newly developed safeguards to state agencies for evaluation.
Furthermore, Anthropic will establish dedicated internal teams focused on joint AI safety research and will allocate significant computing resources to support government-backed testing initiatives.
The company emphasized that transparent, permanent, and robust legal regulations applied equally to all frontier model developers are critical for the future of cyber defense.
America
US Congress nears deal on legislation to ban Chinese-origin cars
The US Congress is closing in on a deal that would ban Chinese-origin cars from entering the US.
Sources speaking to Bloomberg said that Republicans and Democrats from car-making states behind the initiative aim to deliver a bill to President Donald Trump’s desk by the end of the year.
If the law takes effect, the lucrative US car market will be closed to car makers such as BYD and Geely, which have been rapidly gaining market share across Europe and South America with their low-cost, tech-packed electric vehicles.
Chinese cars are already effectively excluded from the US market due to high tariffs imposed by the Department of Commerce and restrictions on Chinese vehicle technology.
Enacting a statutory ban would take these measures a step further to address mounting concerns that the Chinese automotive industry poses “national and economic security threats”.
Details of the measure are still being worked out. According to the sources, Republican Senator Bernie Moreno briefed automotive manufacturer representatives on amendments under consideration to earlier legislation aimed at barring Chinese-origin connected vehicles from entering the US.
These amendments include provisions that would shield Mercedes from a complete ban.
According to the sources, car company representatives generally supported the initiative Moreno outlined, but a Mercedes official expressed reservations, and negotiations remain ongoing.
Sources noted that the bill’s backers are working to pass the measure before the new congressional term begins in January.
In a statement, Mercedes said: “We continue to talk with all stakeholders who share a common interest in seeing this important legislation become law during this session of Congress.”
A bill introduced by a Senate committee earlier this year would have prohibited manufacturers more than 15% owned by Chinese entities from selling vehicles equipped with technology capable of transmitting data.
Mercedes, which has lobbied for changes to this bill, is roughly 10% owned by Chinese state-owned car maker BAIC.
Li Shu Fu, the billionaire founder of Chinese car manufacturer Geely, also owns about 10% of Mercedes shares.
This puts the company above the 15% threshold set in the previous bill.
A source familiar with the talks, speaking on condition of anonymity, said members of Congress were in discussions with multiple parties, including Mercedes, and were optimistic that a deal could be reached this week.
Provisions under discussion include more flexible rules for firms with non-voting, passive shareholdings, as well as how such an approach would be structured and verified.
Republican Senator Katie Britt of Alabama, a key manufacturing hub for Mercedes sport utility vehicles (SUVs), said she was working on the bill with Moreno and expressed confidence that both Mercedes and Alabama would be protected:
“We understand that some adjustments need to be made, but we are confident that Mercedes can continue to thrive in our state and across the nation.”
Banning Chinese-origin cars would also contrast with the more flexible approach adopted by Europe and Canada.
Chinese car manufacturers have gained market share rapidly, particularly in Europe, triggering intense competition with domestic brands.
After Chinese car makers accelerated their market share gains in Europe, the EU began to push back, signing an agreement with China last month to limit imports of hybrid vehicles into the bloc.
America
Elon Musk says he feels pretty stupid after $442m midterm spending
Tesla CEO Elon Musk has said he feels “pretty stupid” for spending $442 million on Republican campaigns after the party suffered heavy losses to Democrats in the 2026 midterm elections.
Speaking remotely at the DealBook Summit in an interview with New York Times columnist Andrew Ross Sorkin, Musk, the world’s richest person, stated that his primary objective had been to help Republicans maintain control of the Senate.
However, Democrats secured the majority, with Democratic candidate Mary Peltola winning the Alaska Senate race to bring the seat distribution to 56 for Democrats against 44 for Republicans.
“I feel pretty stupid spending that much money and not seeing much return on it,” Musk told Sorkin.
Musk added that the outcome could have been far worse without his spending, but noted: “It certainly shows that American elections are not just about money.”
Musk continued: “Clearly you can spend a ton of money on an election and have relatively little impact, which I think was the case here.”
Saying he would spend less if he could turn back the clock, Musk remarked that he did not have much personal interest in the elections because he was focused on growing SpaceX and Tesla.
Musk describes himself as “ideologically moderate”
Musk stated that heading into 2029, he wanted Republicans to control at least one of the House of Representatives, the Senate, or the White House, while wishing that neither Democrats nor Republicans attain a “supermajority”.
“I think we want to have checks and balances here,” Musk said, emphasising that this was his sole motivation.
Describing himself as “ideologically pretty moderate”, Musk said: “I do not think I like any extreme political views.”
Noting during the interview that he was “not MAGA” and was unsure what the term strictly meant, Musk added: “If it means blind adherence to a set of Republican Party policies, I certainly do not have that.”
Rapprochement with Trump and the DOGE era left behind
Musk had previously forged a close relationship with President Trump ahead of the 2024 presidential election, subsequently being appointed to lead the now-defunct Department of Government Efficiency (DOGE).
The pair’s working dynamic led some critics to allege they were operating a “co-presidency”.
However, following a disagreement between the two men, Musk stepped down from his post to focus on his companies, and DOGE closed in July upon the expiration of its mandate under an executive order.
Musk had described himself as “dark gothic MAGA” at an October 2024 Trump rally held at Madison Square Garden.
Critics claimed the font on the black MAGA cap he wore at that rally resembled the Fraktur typeface used by the Nazis during their rule in Germany, though experts later debunked the allegation.
Musk has also backed other right-wing political parties around the world.
In January 2025, he attended a rally of the far-right Alternative for Germany (AfD) party online.
Georgetown University researchers noted that the AfD’s sweeping victory in the German state of Saxony-Anhalt last month raised the possibility of “the first far-right state government in Germany since the Nazi era”.
America
Anthropic IPO set to turn seven co-founders into billionaires
The seven co-founders of Anthropic, the developer behind the Claude artificial intelligence models, are set to join the ranks of multi-billionaires following the company’s initial public offering on the NASDAQ stock exchange.
According to a report by The Wall Street Journal (WSJ), each of the seven founders is expected to make around $250 million from share sales in the IPO, which opened today.
Following the sale, the stake retained by each founder is projected to be worth roughly $18 billion.
The newspaper reported that all the company’s founders belong to millennial generation, with the group including two siblings, three PhD graduates, a former journalist, and a university drop-out.
Amodei siblings left OpenAI to build their own lab
Leading the list is Anthropic’s 43-year-old CEO, Dario Amodei.
Amodei, who holds a doctorate in biophysics, focused on cancer during his early research.
Having previously worked at Baidu and Google, Amodei was invited to OpenAI in 2016 by its CEO, Sam Altman.
Over time, as disagreements arose regarding the hazards tied to AI development, Amodei and six others left OpenAI in 2020 to establish their own laboratory.
Among those set to join the multi-billionaires is Dario’s 39-year-old sister, Daniela Amodei, who serves as Anthropic’s president and board chair.
After graduating from university, Daniela Amodei worked in politics and at the company Stripe. She later joined her brother at OpenAI, leading teams that developed early iterations of its flagship chatbot.
According to the WSJ’s assessment, Daniela Amodei’s greatest strength lies in understanding her brother’s way of thinking and helping to keep the company on the right path.
University drop-out and company “Swiss Army knife” on the list
Another co-founder poised to expand his fortune is 34-year-old Christopher Olah, who heads the company’s interpretability research.
According to Anthropic, Olah, who dropped out of university, has become a “founding figure” in this field.
Tom Brown, Anthropic’s 39-year-old head of computing resources, is also among those set to become multi-billionaires.
Described by the WSJ as a “Swiss Army knife”, Brown co-developed one of the fundamental concepts enabling AI systems to hold dialogue alongside Dario Amodei, negotiates chip deals, and conducts informal lobbying on behalf of the company.
According to the newspaper, Brown also used his connections with the Republican Party to resolve disputes involving SpaceX CEO Elon Musk and the US administration of President Donald Trump that posed threats to the company’s business.
Former journalist and physicists shaped the rules of AI
Anthropic’s fifth co-founder, 38-year-old Jack Clark, transitioned into the AI sector from journalism.
The WSJ noted that Clark, who previously worked as a technology writer at Bloomberg Businessweek, was drawn to the forms of machine learning that paved the way for Claude.
Leaving OpenAI to head Anthropic’s policy division, Clark now leads the Anthropic Institute, which examines the consequences of deploying powerful AI systems.
Another prospective multi-billionaire is theoretical physicist Jared Kaplan, 42, Anthropic’s chief science officer.
Kaplan, who has lectured at Johns Hopkins University since 2012, has scientific interests spanning from quantum gravity to machine learning.
Holding a doctorate in physics from Harvard, Kaplan served as a research consultant at OpenAI in 2019.
Kaplan is also the author of the paper that examined how the capabilities of AI models depend on computing resources, training data volume, and model parameter counts.
Within Anthropic, he continued developing the idea of governing AI systems under a framework of rules designed to keep them from spiralling out of control.
Finally, 36-year-old Sam McCandlish, Anthropic’s chief architect who oversees the parameter development for new models, will also join the ranks of multi-billionaires.
Having worked closely with Kaplan during their time at OpenAI, McCandlish is the co-author of the research paper on scaling laws.
IPO share price reported to potentially reach $83
Anthropic filed for its IPO on 26 October and began trading today on NASDAQ under the ticker ANT.
The IPO is being underwritten by various banks, led by Goldman Sachs and Morgan Stanley.
The company aims to float approximately 825 million shares within a price range of $70 to $78 per share.
Furthermore, according to a report by Bloomberg citing sources, the share price will surpass the designated range and reach roughly $83.
Anthropic is offering 801.4 million shares in the flotation, while existing shareholders plan to sell 23.6 million shares.
According to a company statement, Anthropic and its co-founders aim to raise up to $64.4 billion from the IPO and $26 billion through private sales.
-
America2 weeks agoUS midterms see 97 young winners as Gen Z representation triples
-
Europe2 weeks agoUK and Germany expand military pact with focus on Russian frontier
-
America2 weeks agoNvidia investigates CEO Jensen Huang over link to Chinese spy
-
America2 weeks agoUS secures 35% stake in Venezuelan firm NABEP under 50-year deal
-
Europe2 weeks agoGeorge Szamuely dissects European decline and strategic rifts with US
-
Opinion2 weeks agoThe Mecca Defence Agreement, the Ottoman Empire, and the facts
-
Opinion2 weeks agoSettle, deny, seal: Britain’s formula for colonial torture, from Kenya to Cyprus — 1
-
America1 week agoAbsent Democrats derail floor vote on Pete Hegseth impeachment
