European Union crypto clients should be served through a Markets in Crypto-Assets Regulation (MiCA)-authorized legal entity after the bloc’s July 1 transitional deadline, the European Securities and Markets Authority (ESMA) told Cointelegraph, adding to questions over how global exchanges can keep servicing users in the region.

Crypto asset service providers (CASPs) must hold MiCA authorization to serve clients across the EU and European Economic Area, an ESMA spokesperson told Cointelegraph on Monday.

“EU clients should be serviced through a MiCA-authorized entity,” the ESMA representative said, adding that MiCA protections apply only to the legal entity that is licensed in the EU.

The clarification came shortly after Binance told its users it was adjusting services in certain EU countries, including Poland, France, Spain and Italy, as part of its MiCA transition. Binance said users in other countries would not need to take action if they were not based in a jurisdiction where the exchange operates through a local registered entity, saying in those cases that “no action is required at this time.”

ESMA cites “narrow exemption” for non-EU CASPs

The ESMA said CASPs based outside the EU cannot provide their services to local customers unless they fall under the “narrow exemption” of reverse solicitation provided by Article 61 of MiCA.

Article 61 allows a non-EU crypto company to serve an EU client without a MiCA license only when the client initiates the relationship entirely on their own, without any solicitation, marketing or promotion by the company.

However, the regulation makes clear that the exemption does not apply if a third-country company solicits clients in the EU.

“MiCA established that where a third-country firm solicits clients or prospective clients in the Union […] it shall not be deemed to be a service provided on the client’s own exclusive initiative,” an ESMA spokesperson told Cointelegraph.

Excerpt from ESMA’s list of examples of solicitation by third-country companies. Source: ESMA

The regulator also cited its official solicitation guidelines, which include activities such as operating websites, mobile apps, social media, online advertising, sponsorships and influencer campaigns targeting EU users.

Lawyer questions Binance’s Abu Dhabi servicing model

Screenshots of Binance customer support messages circulating on social media appeared to suggest that some EU users could be serviced through Binance’s Abu Dhabi Global Market entity.

Yuriy Brisov, a lawyer at Digital & Analogue Partners, said an Abu Dhabi license has no effect under MiCA because the jurisdiction is treated as a third country, alongside markets such as the United States or Singapore.

Source: Satoshi Club

“Being regulated in Abu Dhabi does nothing for Binance under MiCA,” Brisov said. “When Binance says some EU users are serviced through the ADGM entity, in MiCA terms that means a non-EU company is serving those users,” he added.

Related: Germany leads MiCA crypto authorization race as Europe’s deadline looms

Brisov said that the reverse solicitation exemption was designed for isolated cases where an EU customer independently approaches a non-EU company, not for maintaining an existing customer base built through years of marketing.

Binance did not respond to repeated Cointelegraph requests for clarification on whether any EU users would be serviced through its ADGM entity after the MiCA deadline.

Magazine: How crypto laws changed in 2025 — and how they’ll change in 2026

A former CEO of Goliath Ventures, Christopher Alexander Delgado, has entered a guilty plea in the U.S. Department of Justice’s cryptocurrency fraud case that prosecutors say relied on investor money to sustain promised returns. The DOJ said the scheme raised at least $400 million from investors through what it described as digital asset “liquidity pools,” with Delgado admitting to wire fraud conspiracy and related offenses.

According to the DOJ, Goliath marketed monthly returns between January 2023 and January 2026. Prosecutors contend that investor funds were not used as represented, but instead were diverted to pay earlier participants, handle withdrawal requests, cover personal luxury spending, and fund business events.

Key takeaways

• Delgado pleaded guilty to conspiracy to commit wire fraud, along with wire fraud and money laundering counts tied to the Goliath Ventures scheme.
• Prosecutors said Goliath promised monthly returns from crypto liquidity pools while using new investor money for withdrawals, payments, and expenses.
• The plea agreement states the scheme resulted in at least $250 million in investor losses and includes forfeiture of extensive luxury assets.
• Sentencing is scheduled for Oct. 8, with Delgado facing substantial prison time if convicted on the fraud and money laundering counts.
• The case has also raised questions about how financial institutions handled Goliath-related transactions, including litigation claims involving major banks and crypto custody.

What the DOJ says Delgado admitted

In a Tuesday announcement, the DOJ said Delgado pleaded guilty to a set of charges stemming from the cryptocurrency investment operation. Prosecutors described a program that promised regular returns using digital asset liquidity pools, spanning January 2023 through January 2026.

Delgado’s plea included admissions that the funds collected from investors were used for purposes inconsistent with the representations made to customers. The DOJ said money from the scheme went toward paying earlier investors and processing withdrawal requests, while also being used for luxury spending and to finance business events.

Under the plea agreement, Delgado admitted that investor losses linked to the scheme totaled at least $250 million. The agreement also requires forfeiture of a portfolio of luxury assets, according to the DOJ.

Forfeiture and sentencing exposure

The DOJ said Delgado agreed to surrender eight properties, 11 vehicles, and more than 50 luxury bags and wallets, along with at least 29 pieces of jewelry. The forfeiture list also includes bank accounts and crypto wallets identified by prosecutors.

From a potential punishment standpoint, the DOJ’s filing indicates that each fraud count carries up to 20 years in prison, while money laundering carries up to 10 years. Delgado’s sentencing has been set for Oct. 8.

Apology on television precedes the plea

Delgado’s guilty plea follows a public apology and television appearance that drew attention to the case. Earlier this year, on May 12, Delgado appeared in an interview with Florida television station WFTV, where he acknowledged he had let investors down and said he had returned to the U.S. and was cooperating with authorities.

In that interview, Delgado claimed that only about $160,000 remained in the company’s bank account at the time of his arrest. He also said other former colleagues were involved in the broader operation.

Scrutiny expands to banks and payment rails

Beyond Delgado’s plea, the Goliath case has drawn scrutiny regarding the financial institutions that processed transactions connected to the scheme. On March 12, investors reportedly filed a proposed class-action lawsuit against JPMorgan Chase, alleging that the bank ignored suspicious activity and allowed Goliath to route investor funds through its accounts.

That lawsuit, as described in earlier coverage, claimed that roughly $253 million moved through a JPMorgan account. It further alleged that about $123 million later transferred to Goliath’s wallets at Coinbase. A separate federal complaint, according to the DOJ, also identified flows involving Bank of America and transfers directly to Coinbase wallets.

For market participants and builders, these allegations underscore a recurring challenge in crypto markets: even when digital assets are involved, traditional banking channels can remain critical bottlenecks for detecting suspicious activity—or, if the system fails, for facilitating fraudulent structures.

With sentencing approaching on Oct. 8, investors and compliance professionals will likely focus on what the court order clarifies about the scheme’s mechanics, the scale of losses, and how payments moved between bank accounts and crypto wallets. The plea does not end the broader questions raised by parallel claims about financial institutions’ role, and future rulings could further shape expectations for safeguards across both traditional and on-chain settlement pathways.

Key Highlights

• FDX shares declined 3.7% during Wednesday’s pre-market session following news of potential $1.4B Supply Chain unit divestiture
• CMA CGM, the globe’s third-largest container shipping operator, is identified as the prospective purchaser
• The company exceeded Q4 FY2026 expectations with earnings of $6.31 per share and revenue of $25.01B, reflecting a 12.5% annual increase
• Company insiders have divested 48,104 shares totaling approximately $17.6M in the previous quarter
• Analyst consensus points to a “Moderate Buy” rating with a mean price objective of $349.60

Shares of FedEx (FDX) experienced a 3.7% decline in Wednesday’s pre-market session, beginning trading at $313.44, following a Financial Times report indicating that negotiations to divest its third-party logistics division have reached an advanced phase.

FedEx Corporation, FDX

The business unit under discussion is FedEx Supply Chain, responsible for warehousing operations, distribution services, and managing product returns. According to reports, CMA CGM, ranked as the world’s third-largest container shipping enterprise, stands as the potential acquirer.

Sources suggest the transaction carries an estimated valuation of approximately $1.4 billion.

This development represents another strategic step in FedEx’s initiative to concentrate on its primary air and ground transportation services. The logistics giant recently separated FedEx Freight (FDXF) into an independent trucking entity.

Divesting the Supply Chain business would further streamline FedEx’s operational structure and intensify its concentration on package delivery services.

Strong Quarterly Performance Amid Strategic Transitions

FedEx delivered impressive Q4 FY2026 financial results on June 23. The shipping giant reported earnings per share of $6.31, surpassing Wall Street projections of $5.91 by $0.40. Total revenue reached $25.01 billion, exceeding the consensus estimate of $24.04 billion and representing a 12.5% increase from the previous year.

FedEx also highlighted enhanced package volume metrics and ongoing cost reductions stemming from its comprehensive transformation initiative.

Stifel’s analyst J. Bruce Chan expressed optimism regarding continued enhancement in profit margins and asset efficiency.

Looking ahead to FY2027, FedEx has provided earnings guidance ranging from $16.90 to $18.10 per share. Wall Street analysts project an average of $20.89 in full-year earnings.

Notwithstanding the earnings outperformance, FDX shares have faced headwinds. The company’s transition to calendar-year financial reporting has introduced some investor ambiguity.

Notable Insider Transactions and Institutional Activity

Regarding insider activity, Executive Vice President Gina F. Adams divested 20,450 FDX shares on April 14 at a price of $366.45 per share, generating approximately $7.49 million in proceeds. This transaction reduced her holdings by 51%.

Board member Susan C. Schwab sold 5,795 shares the following day at $369.00 per share, totaling roughly $2.14 million — representing a 36.6% decrease in her ownership stake.

Cumulatively, company insiders have disposed of 48,104 FDX shares valued at approximately $17.6 million during the past 90-day period.

Institutional investor Janney Montgomery Scott reduced its FedEx holdings by 16.6% during Q1, selling 10,993 shares. The firm maintains ownership of 55,268 shares with an estimated value of $19.7 million.

Institutional investors control 84.47% of FedEx stock.

Wall Street analysts have recently adjusted their price projections. Truist reduced its target from $425 to $365 while retaining a Buy recommendation. Stifel decreased its objective from $442 to $326, also maintaining Buy. HSBC lowered its forecast to $289.89 with a Hold rating. Sanford C. Bernstein established a $397 target.

Morgan Stanley maintains the sole Underweight rating, projecting a $230 price target.

The prevailing Wall Street consensus stands at Moderate Buy, featuring an average price objective of $349.60.

FedEx’s 52-week trading range extends from a minimum of $172.88 to a maximum of $345.36. The equity’s 50-day moving average currently rests at $361.08.

A quarterly dividend distribution of $1.22 per share was issued on July 7, yielding 1.6%.

Artificial intelligence company Anthropic is set to restore public access to its most powerful AI models, Claude Fable 5 and Mythos 5, weeks after they were pulled offline under a directive from the US government. 

Anthropic’s two latest models have been restricted from public access since June 12, when the government applied export controls following a report in which researchers bypassed Fable 5’s safeguards, forcing Anthropic to pull all access to the models immediately. The government lifted those restrictions on Wednesday, stated Anthropic. 

“After a series of productive conversations with the US government, we’re redeploying the model with a new set of classifiers to target and block more cybersecurity tasks,” Anthropic said. 

The suspension of the models raised concerns about state control over frontier AI technology and set a dangerous precedent, according to experts and technologists. The export controls also highlighted White House concerns about a potential national cybersecurity threat if these powerful models were jailbroken and used for malicious purposes. 

Getting best tech deployed remains priority 

US Secretary of Commerce Howard Lutnick said on X on Wednesday, “Over the past two weeks, we have worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the US Government and to strengthen America’s leadership in AI.” 

Meanwhile, White House Chief of Staff Susie Wiles said on X that government priority remains to “get the best [AI] tech deployed as quickly and safely as possible.”

Related: AI researcher claims he’s already bypassed Anthropic’s Fable 5 guardrails

The restrictions came after the government became aware of a report in which Amazon researchers found a method of bypassing Fable 5’s safeguards, prompting the model to identify several software vulnerabilities. 

In a blog post, Anthropic argued this wasn’t a risk unique to Fable 5, as weaker models could also identify the same vulnerabilities and produce the same exploit.

AI jailbreak classifications proposed 

Anthropic has also begun drafting a consensus framework with Amazon, Microsoft, Google and other partners in its Project Glasswing — a collaboration announced in April to safeguard against AI cybersecurity threats — for “assessing the severity of AI jailbreaks.”

Anthropic’s cybersecurity safety classifiers and how jailbreaks interact with safety classifiers. Source: Anthropic

The company is also scaling up collaboration with the US government on AI model testing and safeguards. “This will include pre-release access to models and safeguards for evaluation, information sharing on jailbreaks and misuse, and dedicated resources for joint research,” it stated.

A well-known AI researcher claimed to have jailbroken Fable 5 within 48 hours of its launch in June, before the government restrictions, and shared screenshots showing how he bypassed the model’s safety guardrails.  

Magazine: AI is banking the unbanked in Africa… faster than crypto

Key Highlights

• Bloom Energy shares rose 10% during after-hours trading on July 1, 2026.
• Brookfield Asset Management expanded its financing pledge for Bloom-powered AI infrastructure from $5 billion to $25 billion.
• Capital originates from Brookfield’s AI Infrastructure Fund, focused on worldwide deployment of Bloom’s solid oxide fuel cell technology.
• Bloom’s current 2026 project roster features partnerships with Oracle and Nebius.
• BMO Capital maintained its Market Perform stance with a $279 target price, emphasizing the arrangement doesn’t translate to instant confirmed orders.

Bloom Energy shares experienced a 10% spike in extended trading hours on July 1 following the announcement that Brookfield Asset Management would dramatically expand their collaborative AI data center power initiative. Prior to the after-hours movement, shares were hovering near $302.70.

Bloom Energy Corporation, BE

The partnership revealed that Brookfield plans to increase its financial backing for Bloom-equipped infrastructure from $5 billion to $25 billion — representing a 400% increase since the original agreement was established in late 2025.

Capital will be sourced from Brookfield’s AI Infrastructure Fund, designed to facilitate worldwide implementation of Bloom’s on-premises solid oxide fuel cell technology for data center applications.

This substantial expansion underscores accelerating requirements from hyperscale operators and AI developers seeking rapid, dependable power solutions that minimize grid dependency. Fuel cell technology delivers continuous, on-location energy generation — an increasingly attractive feature as AI computational demands escalate.

The partnership structure positions Brookfield as the financial provider with extensive global capabilities while Bloom furnishes rapidly deployable fuel cell systems. Their combined approach integrates power generation infrastructure directly into data center architecture.

Bloom’s Chief Commercial Officer Aman Joshi noted the increased commitment reflects the market’s accelerating pace, highlighting numerous large-scale AI power transactions. Brookfield’s AI Infrastructure division head Sikander Rashid characterized the expansion as reinforcing Brookfield’s position among leading global AI infrastructure investors, delivering comprehensive “end-to-end solutions, from electrons to tokens.”

Bloom’s Active 2026 Portfolio

The Brookfield partnership extension represents just one component of Bloom’s robust 2026 activity. The organization has secured a substantial agreement with Oracle and finalized a fresh partnership with Nebius — contributing to an increasingly active project pipeline.

Brookfield’s AI Infrastructure Fund, established in 2025, targets deployment of $100 billion across AI facilities, energy systems, computing infrastructure, and associated capital investments. The company maintains over $100 billion in existing digital infrastructure and renewable energy investments.

BMO Capital clarified the $25 billion commitment shouldn’t be interpreted as confirmed immediate backlog. Rather, it establishes a programmatic financing framework that reduces capital barriers for clients and may accelerate implementation timelines.

BMO presented three potential demand projections for Bloom, spanning 2.4 gigawatts to beyond 5.0 gigawatts of possible deployment, contingent on structural arrangements, tax considerations, and revenue composition.

Analyst Perspectives

BMO confirmed its Market Perform assessment and $279 target price on July 1. While recognizing the deal demonstrates increasing institutional trust in Bloom’s position within AI infrastructure, the firm maintained its current rating given the nascent stage of large-scale solid oxide fuel cell implementation.

The Street’s overall consensus stands at Moderate Buy, reflecting nine Buy recommendations and 10 Hold ratings. The mean price target reaches $269.42.

Bernstein SocGen Group initiated coverage recently with a Market Perform designation and $276 price objective. UBS maintains a Buy recommendation, referencing regulatory modifications enabling faster connection of major energy consumers to the national transmission infrastructure.

Bloom’s board additionally authorized a performance-linked restricted stock unit award for CEO Dr. KR Sridhar, with vesting conditions connected to revenue milestones spanning July 1, 2026, through December 31, 2029.

US-listed spot Bitcoin exchange-traded funds (ETFs) posted a record $4.5 billion in net outflows in June, more than three times the $1.25 billion Strategy is authorized to raise through its new Bitcoin monetization program.

The record monthly withdrawals pushed US spot Bitcoin ETFs to roughly $5.5 billion in year-to-date net outflows for 2026, reducing cumulative net inflows since the funds launched to about $51.2 billion, according to SoSoValue data updated on Wednesday.

BlackRock’s iShares Bitcoin Trust (IBIT) accounted for about 79% of June’s withdrawals, posting $3.55 billion in net outflows, according to Farside Investors.

Monthly flows in US-listed spot Bitcoin ETFs. Source: SoSoValue

The figures highlight weakening demand for US spot Bitcoin ETFs, despite much of the market’s attention remaining fixed on developments surrounding the industry’s largest corporate Bitcoin treasury company.

Bitcoin ETF holdings fall below year-ago levels despite higher inflows

According to SoSoValue, cumulative net inflows into US spot Bitcoin ETFs have risen 4.6% from about $49 billion a year earlier. But CryptoQuant data shows the funds now hold less Bitcoin than they did at the same time last year.

“US-based Bitcoin ETF holdings are now lower than at this same day last year,” CryptoQuant’s head of research Julio Moreno wrote on X on Tuesday.

Source: Julio Moreno

Moreno said overall demand for Bitcoin continues to weaken, with total holdings across US spot Bitcoin ETFs falling below 1.25 million BTC.

Related: Swan’s Cory Klippsten sees record Bitcoin holder supply revealing early bottom

ETF withdrawals dwarf Strategy’s Bitcoin plan

Strategy announced its Bitcoin monetization program on Monday as part of a broader capital framework designed to support dividend obligations tied to its preferred securities, a move widely viewed by investors as a response to growing funding pressure within the company’s structure.

Source: Jeff Dorman

The move drew mixed reactions across the community, with some viewing it as financial flexibility while others flagged concerns over the new capital structure’s long-term sustainability and argued it could ultimately sell much more than $1.25 billion.

Strategy’s Class A common stock (MSTR) initially surged as much as 12% to above $90 following Monday’s announcement before reversing course and closing at $86.93 on Tuesday, down 6.2% on the day, according to Yahoo Finance.

Meanwhile, Strategy’s preferred stock (STRC) traded higher at $84.86 on Tuesday, according to Yahoo Finance.

Magazine: Bitcoin slides to $58K, XRP hits $1 but onchain data promising: Market Moves

Key Takeaways

• Amazon (AMZN) shares started Wednesday’s session at $238.51, maintaining a $2.57 trillion market capitalization.
• Futurum Equities analyst Shay Boloor projects Amazon could reach $1 trillion in yearly revenue by 2028, a first for any corporation.
• The tech giant has earmarked approximately $200 billion for infrastructure investments in 2026, supported by OpenAI agreements exceeding $100 billion.
• Revenue from Amazon’s proprietary semiconductor offerings, Graviton and Trainium, has surpassed a $20 billion annual run rate.
• Legal challenges include a $2.25 million Federal Trade Commission settlement and fresh litigation in Australia concerning Prime Video advertising.

Amazon (AMZN) kicked off Wednesday’s trading at $238.51 per share. With a substantial market capitalization of $2.57 trillion, the stock currently trades within its 52-week range of $196.00 to $278.56.

Amazon.com, Inc., AMZN

According to one market observer, the e-commerce and cloud computing behemoth has only scratched the surface of its potential. Shay Boloor, Chief Market Strategist at Futurum Equities, believes Amazon is “on track to become the first company to cross $1 trillion in annual revenue by 2028.”

Boloor highlighted Amazon’s diversified operations spanning e-commerce, cloud computing, logistics networks, advertising platforms, and artificial intelligence capabilities. He characterized the company as “one of the most important infrastructure companies in the world.”

Massive Capital Allocation for AI Infrastructure

Amazon has outlined plans to allocate approximately $200 billion toward capital investments throughout 2026. A significant portion of this expenditure stems from substantial client agreements, notably commitments exceeding $100 billion from OpenAI.

The company’s proprietary semiconductor portfolio, featuring Graviton and Trainium processors, has achieved a $20 billion annualized revenue milestone. Amazon reports this division is experiencing growth rates exceeding 100%.

These in-house chip solutions are anticipated to reduce operational expenses and enhance profit margins over the coming years. This strategic advantage holds particular significance for an organization managing vast data center infrastructure globally.

Amazon surpassed Wall Street expectations with its first-quarter earnings released this past April. The company generated $181.52 billion in quarterly revenue, exceeding analyst projections of $177.30 billion.

Per-share earnings reached $2.78, substantially outperforming the consensus forecast of $1.66 per share.

Institutional investment activity reflects confidence in Amazon’s trajectory. Cardinal Point Capital Management ULC expanded its Amazon holdings by 13.6% during the first quarter, purchasing an additional 4,450 shares to reach a total position of 37,124 shares valued at approximately $7.73 million.

Numerous other investment firms executed comparable transactions. Brighton Jones LLC increased its stake by 10.9%, now controlling more than 4 million shares with an estimated value approaching $885 million.

Institutional investors collectively control 72.2% of Amazon’s outstanding shares, representing substantial ownership concentration.

Executive Stock Sales Continue

Despite positive fundamentals, corporate insiders have been reducing positions. CEO Matthew Garman divested 15,467 shares in May at $263.40 per share, generating proceeds exceeding $4 million.

SVP David Zapolsky similarly reduced his holdings, selling 9,270 shares at $268.53 apiece. Throughout the past three months, company insiders have collectively sold approximately $51.4 million in stock.

Amazon continues navigating regulatory challenges as well. The corporation reached a $2.25 million settlement agreement with the FTC, while simultaneously confronting new legal action in Australia regarding Prime Video advertising practices.

On a more positive note, AWS recently unveiled a $1 billion Forward Deployed Engineering initiative. This program aims to position AI specialists directly within customer organizations to accelerate enterprise-level implementation.

A recently published Jefferies survey revealed 95% of information technology decision-makers intend to expand cloud infrastructure budgets in 2026. Industry analysts identify AWS as a primary beneficiary of this spending trend.

Wall Street’s price projections demonstrate considerable bullishness. Stifel Nicolaus established a $319 price target, while Susquehanna elevated its forecast to $325 accompanied by a “positive” outlook.

Currently, 57 analysts maintain Buy recommendations on Amazon shares, compared to just three Hold ratings. MarketBeat reports a consensus “Moderate Buy” designation with an average price objective of $312.78.

Technical indicators show Amazon’s 50-day moving average at $255.10, while the 200-day average stands at $234.31. The stock trades with a price-to-earnings ratio of 28.53 and exhibits a beta coefficient of 1.44.

Anthropic has moved to restore public access to its most capable AI models, Claude Fable 5 and Mythos 5, after the US government ordered their temporary shutdown over cybersecurity concerns. The models were taken offline in mid-June following reports that researchers found ways to bypass safeguards, prompting Anthropic to suspend public redeployment while controls were tightened.

According to posts from Anthropic on Wednesday, the government lifted the restrictions and the company is redeploying Fable 5 with an updated safety approach. Anthropic said it is using “a new set of classifiers” designed to better target and block more cybersecurity tasks, reflecting a shift from broad access to narrower, more selectively controlled operation.

Key takeaways

• Anthropic says US export-related restrictions on Claude Fable 5 and Mythos 5 were lifted on Wednesday, enabling renewed public access.
• The models were pulled after reports of bypassing Fable 5’s safeguards that could potentially help identify vulnerabilities for malicious use.
• Anthropic plans to redeploy with revised cybersecurity safety classifiers aimed at blocking additional harmful task types.
• US officials emphasized “safe” deployment while reviewing model alignment with government priorities.
• Anthropic is also working on a broader framework to assess jailbreak severity through its Project Glasswing partnership network.

Why Anthropic’s access was paused

Public access to Anthropic’s two latest flagship models—Claude Fable 5 and Mythos 5—was restricted starting June 12. The immediate trigger, Anthropic indicated, was an export-control directive tied to government review of a report describing how researchers could bypass safeguards on Fable 5.

In that report, the bypass method allegedly allowed the model to surface multiple software vulnerabilities. Anthropic responded by pulling access to the models rather than leaving the public-facing system in a state the government considered too risky. The broader implication was that frontier model capabilities, when paired with inadequate resistance to misuse, could translate quickly into real-world security threats.

What changed in the redeployed model

Anthropic said the suspension is ending after “productive conversations with the US government.” In its statement, the company framed the update as an operational safety upgrade rather than a complete redesign: the redeployed model will use “a new set of classifiers” intended to detect and block more cybersecurity tasks.

This matters for users and developers because classifier-based controls directly affect what kinds of requests the model will refuse or redirect. In practical terms, those controls can determine whether legitimate security analysis workflows remain usable while high-risk directions are filtered more aggressively.

Anthropic also addressed a key concern raised during the shutdown: the company argued that the issue wasn’t uniquely tied to Fable 5. In a blog post titled “Redeploying Fable 5,” Anthropic said weaker models could potentially identify similar vulnerabilities and produce the same exploit pathways. That framing suggests the company views the problem as a category risk across model capability levels, not a single-model anomaly.

US government concerns and the push for rapid deployment

US Secretary of Commerce Howard Lutnick said on X that, over the past two weeks, the government worked closely with Anthropic to “analyze and approve Fable 5,” aiming for alignment across the US government and to strengthen American AI leadership. Separately, White House Chief of Staff Susie Wiles said the priority remained to “get the best [AI] tech deployed as quickly and safely as possible.”

While those remarks do not change the underlying technical issues, they highlight the policy balance driving the intervention: the US government wants frontier models available, but only under conditions it believes reduce the chance of misuse—particularly misuse involving cybersecurity.

The White House concern, according to the article’s context, centered on the possibility of jailbreaking: if powerful models can be coerced into producing harmful instructions, they could become a national security risk. That is the logic behind export-control-style restrictions for models deemed high-impact and harder to contain once broadly available.

Anthropic’s Project Glasswing and a new approach to jailbreak severity

The model shutdown also cast attention on how AI safety should be tested and governed beyond one-off fixes. In the wake of the restrictions, Anthropic said it has begun drafting a consensus framework with partners including Amazon, Microsoft, Google and others for assessing the severity of AI jailbreaks.

The effort is connected to Project Glasswing, a collaboration announced in April focused on safeguarding against AI cybersecurity threats. A key goal appears to be establishing a shared way to evaluate jailbreak attempts: not just whether a safety system can be bypassed, but how dangerous the bypass outcome is and how consistently it can be reproduced.

Anthropic also said it is scaling up collaboration with the US government on model testing and safeguards. The company described plans that include pre-release access to models and safeguards for evaluation, information sharing about jailbreaks and misuse, and dedicated resources for joint research.

In parallel, an AI researcher previously claimed to have jailbroken Fable 5 within 48 hours of its June launch, before the government restrictions were applied. The researcher shared screenshots showing how the guardrails were allegedly bypassed. While that claim is reported as part of the surrounding context, it underscores why governments and model providers view guardrail testing as an ongoing race rather than a one-time checkpoint.

For the broader AI market, these developments raise a familiar tension: the same capabilities that make frontier models useful also raise the cost of safety failures, especially when cybersecurity misuse is possible. The redeployment suggests Anthropic believes the controls can be improved quickly enough to keep pace, but the existence of an industry-wide framework proposal indicates the challenge is bigger than one company’s deployment decisions.

Going forward, readers should watch for how Anthropic’s classifier changes alter real user behavior—what kinds of cybersecurity requests remain available, which patterns are newly blocked, and whether the proposed jailbreak-severity framework results in more transparent testing standards across major model providers.