Hugging Face is investigating a security incident after its production infrastructure was compromised in an intrusion the company says involved an autonomous AI agent, raising fresh concerns about how artificial intelligence could reshape offensive cyber operations.
In a security disclosure published on July 16, the open-source AI platform said the attack leveraged an autonomous agent framework built on top of an agentic security research environment powered by a large language model (LLM). According to the company, the system executed thousands of actions across multiple sandboxed environments, allowing the attackers to move through internal infrastructure and obtain unauthorized access to datasets and service credentials.
The company said the intrusion began when a malicious dataset exploited two separate code execution paths on a processing worker. After establishing an initial foothold, the attacker reportedly escalated privileges to node-level access before collecting cloud and cluster credentials and moving laterally into several internal clusters.
Hugging Face has not yet confirmed whether customer or partner information was affected and said its investigation remains ongoing.
The incident has attracted attention across the cybersecurity community because it suggests that AI systems may now be capable of carrying out increasingly complex intrusion workflows with limited human intervention. Unlike traditional automated malware or scripts that perform predefined tasks, autonomous AI agents can adapt to changing environments, plan sequences of actions and make decisions throughout an attack.
Researchers have long warned that advances in generative AI could lower the barrier for sophisticated cyberattacks by accelerating vulnerability discovery, reconnaissance, privilege escalation and post-compromise activities. While many of these scenarios have remained largely theoretical, Hugging Face's disclosure indicates that elements of these capabilities may already be appearing in real-world operations.
According to the company's investigation, the attacking system generated thousands of individual actions during the compromise, demonstrating a level of operational scale that would normally require substantial manual effort.
Hugging Face co-founder and CEO Clément Delangue said the incident reinforces the view that threat actors are already adopting AI agents in offensive operations. He also argued that restricting advanced AI models behind commercial APIs alone is unlikely to prevent misuse because determined attackers can often circumvent safety controls, while defenders may lose valuable access to tools needed for security research and incident response.
The company encountered another challenge during its investigation when content moderation mechanisms on a frontier AI model reportedly prevented analysts from processing portions of the attack data. To continue the forensic investigation, the security team instead relied on GLM-5.2, an open-weight language model that was deployed within Hugging Face's own infrastructure.
Using the model, investigators reconstructed the attack timeline, identified indicators of compromise, mapped affected credentials and accelerated forensic analysis that would otherwise have required significantly more manual effort. The company also revoked compromised credentials, rotated authentication tokens and remediated the exploited vulnerability.
Security researchers say the incident highlights both the opportunities and limitations of AI-assisted security operations. While AI can substantially reduce investigation time by processing large volumes of telemetry, organizations may encounter operational constraints if externally hosted models refuse to analyze sensitive security artifacts because of built-in safety guardrails.
Industry experts increasingly argue that enterprises should maintain trusted self-hosted AI models that can support internal incident response without exposing sensitive forensic data to external services.
The disclosure comes amid bigger concerns about the growing availability of permissive AI models that operate with fewer content restrictions. Recent threat intelligence research has identified thousands of publicly accessible models advertised as uncensored or unrestricted, raising concerns that malicious actors have expanding access to AI systems capable of assisting offensive cyber activities.
Cybersecurity professionals caution that AI is changing the economics of cybercrime by enabling attackers to automate portions of reconnaissance, exploitation, credential harvesting and post-compromise operations. As these technologies continue to mature, sophisticated attack capabilities may become accessible to a broader range of threat actors.
For defenders, the incident reinforces the need to integrate AI into security operations rather than relying solely on conventional manual workflows. AI-assisted detection, forensic analysis and incident response are increasingly becoming essential capabilities as organizations attempt to match the speed and scale of modern attacks.
Although the investigation into the Hugging Face breach remains ongoing, the incident serves as another indication that autonomous AI systems are beginning to influence both offensive and defensive cybersecurity strategies. As organizations continue adopting AI throughout their technology environments, security teams will need to prepare for a future in which machine-speed attacks are met with equally intelligent defensive capabilities.
The cyberattack involving Ernst & Young (EY) has entered a new phase after the ShinyHunters extortion group claimed responsibility for the intrusion, alleging that it stole data from the third-party support ticket platform used by the global professional services firm. While EY has acknowledged the underlying breach, the company has not confirmed the group's claims, and no leaked data has been independently verified at the time of writing.
The development comes weeks after EY disclosed that an unauthorized party had accessed a third-party support ticket platform used by its IT teams between March 28 and April 12, 2026, potentially exposing documents associated with client tax preparation. The firm had previously informed affected customers that files stored within the support environment could contain personal and financial information submitted through IT support requests, along with documents related to tax services.
According to breach notifications filed with several U.S. state regulators, the exposed records may include sensitive information such as client names, addresses, Social Security numbers, financial account details, payment card information, and other tax-related records. However, EY has not disclosed how many individuals were affected, whether customers outside the United States were impacted, or the identity of the third-party support platform involved in the incident.
The latest development centers on ShinyHunters' public assertion that it carried out the attack and obtained data from the compromised environment. The group has reportedly threatened to publish the allegedly stolen information if its demands are not met. Despite these claims, EY has neither attributed the incident to ShinyHunters nor confirmed that the attackers possess company or client data. Security researchers also note that threat actors have, on occasion, exaggerated or misrepresented claims to increase pressure on victims, making independent verification essential before drawing conclusions.
EY has stated that it immediately activated its incident response procedures after detecting suspicious activity and engaged an independent cybersecurity firm to assist with forensic analysis and remediation. The company says it has contained the unauthorized access, secured the affected environment, and notified relevant federal law enforcement authorities. It also maintains that its investigation has found no evidence that the compromised information has been misused or that individual clients were specifically targeted.
As part of its response, EY continues to provide eligible affected customers with 24 months of complimentary Experian identity restoration, identity monitoring, and credit monitoring services, with enrollment available through October 31, 2026.
Although the breach itself has already been disclosed, the emergence of an alleged threat actor highlights how cyber incidents often evolve long after the initial discovery. Public claims made by ransomware or extortion groups can influence regulatory scrutiny, customer communication, and incident response strategies, even before their assertions are independently verified.
The incident also reinforces the importance of third-party risk management. Organizations that rely on external platforms to process or store sensitive customer information should continuously assess vendor security controls, restrict unnecessary access to confidential data, and maintain comprehensive monitoring and incident response capabilities to reduce the impact of supply chain compromises.
Nearly seven million people are being notified after a cyberattack on Atlanta-based auto insurer AssuranceAmerica exposed highly sensitive personal information, including driver's license numbers, Social Security numbers and insurance records, raising concerns about long-term identity theft risks.
According to the company's breach notice and filings submitted to state regulators, the incident began on March 16, 2026, when a threat actor gained unauthorized access to AssuranceAmerica's internal network using compromised employee credentials obtained through a phishing attack. The company detected suspicious activity the following day, secured the affected systems, and launched a forensic investigation to determine the scope of the compromise.
The investigation later revealed that the attackers had copied files containing personal information belonging to approximately 6.99 million individuals. The exposed data varies by person but may include names, residential addresses, driver's license numbers, Social Security numbers, taxpayer identification numbers, insurance policy and account details, claims information, as well as driver and vehicle records.
The scale of the breach makes it one of the larger disclosures involving government-issued identity documents this year. South Carolina alone reported that 611,046 residents may have been affected, according to the state's Department of Consumer Affairs.
Unlike passwords, driver's license numbers are not easily replaced after they are exposed. These identifiers are widely used to verify identity across banks, insurers, vehicle rental companies, government agencies and financial institutions. When combined with Social Security numbers and other personally identifiable information, they can enable criminals to apply for loans, open fraudulent accounts, submit false tax returns or impersonate victims during identity verification processes.
Law firm Edelson Lechtzin LLP, which announced an investigation into the incident, warned that the compromised information could be used to facilitate identity theft and other forms of financial fraud.
Although AssuranceAmerica identified the intrusion within roughly 24 hours, affected individuals were not notified until late June after investigators completed their review of the compromised data on June 15. The nearly three-month gap between the initial breach and customer notifications has drawn attention to the time required to determine exactly whose information had been accessed before notifications could be issued.
In its public notice, AssuranceAmerica said it disabled the compromised accounts, reset credentials, strengthened network monitoring and provided additional cybersecurity awareness training to employees. The company also engaged external forensic specialists to investigate the incident. However, it has not publicly confirmed whether all affected individuals will receive complimentary credit monitoring or identity protection services.
The AssuranceAmerica breach comes amid a growing number of incidents involving government-issued identity documents. In June, Texas disclosed a separate cyberattack affecting approximately three million driver's license and passport records maintained by the Texas Parks and Wildlife Department, adding to a broader trend of organizations reporting the theft of sensitive identification data.
The growing reliance on digital identity verification has also increased the amount of personal identification collected by businesses and online platforms. As governments and private organizations increasingly require users to upload driver's licenses and other official documents for account verification and age checks, cybersecurity experts warn that breaches involving these records can have lasting consequences because many of these identifiers cannot be easily changed once exposed.
Individuals who may have been affected are encouraged to closely review financial and insurance accounts for suspicious activity, consider placing a credit freeze or fraud alert with the major credit bureaus, monitor their credit reports for unauthorized accounts and remain cautious of phishing emails or phone calls that attempt to exploit information exposed during the breach. Victims should also follow guidance issued by their state consumer protection agencies and promptly report any suspected identity theft.
The Coca-Cola Company has revealed that a ransomware attack targeting its Fairlife dairy business has temporarily disrupted production across the United States after threat actors gained unauthorized access to company systems, including those supporting manufacturing operations.
The incident was disclosed in a Form 8-K filing with the U.S. Securities and Exchange Commission (SEC), a regulatory filing used by publicly traded companies to report significant corporate events. According to Coca-Cola, the cyberattack affected certain Fairlife systems, including production-related infrastructure, prompting the company to temporarily suspend manufacturing at its U.S. facilities while recovery efforts are underway.
Upon detecting the unauthorized activity, Coca-Cola said it immediately activated its incident response and business continuity protocols to contain the incident and minimize operational disruption. The company has engaged external cybersecurity advisors and experts to support its investigation and recovery efforts, while law enforcement has also been notified.
Although manufacturing operations have been interrupted, Coca-Cola emphasized that the ransomware attack has not affected the quality or safety of Fairlife products. The temporary production halt is part of the company's response as it works to restore impacted systems and verify operational readiness before resuming normal manufacturing activities. Fairlife's Canadian production facilities continue to operate normally and have not been affected by the incident.
The company said its investigation remains ongoing and that it is continuing to assess both the nature of the attack and its potential business impact. At this stage, Coca-Cola has not determined whether the incident is reasonably likely to have a material effect on the company's financial condition or overall operations.
Fairlife is one of Coca-Cola's dairy brands and manufactures a range of ultra-filtered milk products, protein shakes and nutrition beverages sold across the United States. Its product portfolio includes Ultra-Filtered Milk, Core Power Protein Shakes and Nutrition Plan.
Several aspects of the incident remain undisclosed. Coca-Cola has not confirmed whether attackers exfiltrated any data during the intrusion, whether the company has received an extortion demand or which ransomware operation may be responsible for the attack. As of publication, no known ransomware group has publicly claimed responsibility for the incident.
Ransomware attacks increasingly target organizations' operational environments in addition to traditional corporate networks, as disrupting production can exponentially multiply pressure on victims during recovery efforts. Many modern ransomware operations also employ double-extortion tactics by stealing sensitive information before encrypting systems and later threatening to publish the stolen data unless a ransom is paid. However, Coca-Cola has not indicated that any data theft occurred in this incident, and there is currently no public evidence confirming that attackers exfiltrated information from Fairlife's systems.
When asked whether data had been stolen, whether the company had received an extortion demand or which ransomware group may have been behind the attack, a Coca-Cola spokesperson declined to provide additional details beyond the company's public statement.
Coca-Cola continues to restore affected systems while its investigation remains ongoing, with U.S. Fairlife production expected to resume once recovery efforts are completed and manufacturing systems have been safely brought back online.
Mount Royal University (MRU) has confirmed that threat actors stole data and deleted files after breaching the university's network in a cyberattack that continues to affect recovery efforts weeks after the incident.
In an update published on its website, the Calgary-based public university said the attack occurred on June 17 and that internal technical teams are working alongside external cybersecurity specialists to investigate the intrusion, determine its full scope, and restore affected systems.
The cyberattack disrupted a wide range of university services, including internet connectivity, online platforms, and several internal systems used across campus. Recovery efforts remain ongoing, with the university warning that restoring all affected services may take several weeks or, in some cases, months.
According to the university's investigation, attackers gained unauthorized access to data stored on the institution's "H drive," a file storage system used by students and employees. Investigators have confirmed that files stored within certain folders were accessed and exfiltrated before the attackers deleted the original copies, a move that has further complicated recovery operations.
"We regret to inform our community that our investigation has now shown that data within certain folders on the University's 'H drive' was accessed and taken by an unauthorized actor," the university said in its advisory.
MRU said the affected folders contained information relating to current and former students, current and former employees, as well as other individuals whose data was stored within the impacted environment. The university has not yet disclosed the exact categories of information exposed or the total number of people affected.
The investigation also found that attackers deleted data stored on a separate departmental file storage system known as the "J drive." While the university said there is currently no evidence that information from the J drive was accessed or copied before it was erased, officials cautioned that recovering the deleted data remains an ongoing process and acknowledged that a complete restoration may not be possible.
The university has reported the incident to the Alberta Information and Privacy Commissioner and notified law enforcement authorities. Officials added that determining the precise impact for each affected individual will take time because the deletion of files has made forensic analysis more complex. Individuals whose information is confirmed to have been affected will receive direct notifications as the investigation progresses.
Responsibility for the attack has been claimed by the cybercrime group CMD Organization, which has published samples of what it alleges is stolen university data, including passport scans and other sensitive documents.
The group is demanding a ransom of 30 Bitcoin, valued at approximately $1.9 million at current exchange rates, and has reportedly given the university six days to respond before releasing additional data. CMD Organization also appears to operate an auction-based extortion model, advertising exclusive access to stolen datasets for the highest bidder through both clear web and dark web leak sites. At the time of writing, the group lists approximately 30 organizations on its extortion portal.
Founded more than a century ago, Mount Royal University currently serves about 11,560 students, including roughly 12,500 undergraduate learners.
As recovery work continues, the university said it will provide additional updates as more information becomes available. MRU is also offering two years of credit monitoring and identity theft protection to current employees and individuals who have worked at the university within the past five years.
A large-scale password spraying campaign targeting Microsoft 365 environments through Microsoft’s Azure Command-Line Interface (Azure CLI) generated more than 81 million authentication attempts and compromised at least 78 user accounts across 64 organizations, according to cybersecurity firm Huntress.
Huntress said the activity was observed between June 12 and June 21, with attackers typically compromising two to four accounts per day before activity surged around June 22, when 23 organizations were affected. Most of the login attempts originated from AS32167, an autonomous system associated with hosting provider LSHIY LLC.
The company said the campaign formed part of a larger wave of credential-spraying attacks spanning multiple autonomous systems and noted that the volume of such attacks across its customer base has increased more than 155-fold during the past six months. Investigators believe the operation relied primarily on previously exposed username-and-password combinations obtained from credential leak collections.
A key element of the campaign was the use of the OAuth Resource Owner Password Credentials (ROPC) flow through Azure CLI. Although ROPC has been deprecated in OAuth 2.1, it can still exchange valid usernames and passwords directly for access tokens without an interactive sign-in prompt. Huntress said this allowed attackers to authenticate successfully in environments where multi-factor authentication policies did not fully cover that authentication flow.
The investigation identified several configuration gaps among affected organizations, including MFA policies applied only to certain cloud applications or user groups, enforcement limited to non-trusted locations, and policies that had been configured but never enforced. Huntress also found that eight impacted organizations had no MFA policy enabled.
Huntress emphasized that the findings should not be interpreted as evidence that MFA is ineffective. Instead, organizations should review Conditional Access policies, eliminate deprecated authentication methods where possible, ensure MFA protections apply to all supported sign-in flows, and monitor Azure CLI authentication activity for unusual login patterns.
The IPv6 address range used in the campaign belongs to LSHIY, an internet infrastructure provider registered in Hong Kong, Wuhan, China, and New York. Huntress said it reported the activity through the provider’s abuse-reporting channel but had not received a response.