October 1, 2026

What happened

Google announced Gemini 4 Argon and is initially rolling it out to trusted cyber defenders through its Fairwind Program. [1]

Google says Argon can locate and validate critical software vulnerabilities and patch them without human help. [2]

Google says it will provide a version without cyber guardrails to the participating defenders and its internal teams. [2]

Why it matters

The reported rollout is limited initially to trusted cyber defenders and Google’s internal teams; developers, enterprises, and consumers are slated to receive Argon later, beginning with paid API customers. [1] [2] [3]

Known limitations

The supplied evidence does not provide independent validation, technical details, vulnerability examples, or information about the model’s patching conditions and results. [1] [2] [3]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   Google announced Gemini 4 Argon, its new frontier AI model, and is rolling it out to a set of trusted cyber defenders through its Fairwind Program.

2. [2]

   Google says the model can locate critical software vulnerabilities, validate them, and patch them without human help, and it will release a version without cyber guardrails to those defenders and to its own internal teams.

3. [3]

   Developers, enterprises, and consumers get Argon later, starting with paid API customers and Google … More → The post Google says Gemini 4 Argon can find and patch critical software flaws appeared first on Help Net Security .

Read the original article →

What happened

Reported employment-scam victims more than tripled over 12 months across more than 370 banks and other financial institutions in 21 countries. [1]

The reported 258% increase exceeded every other scam type, while total reported scams across the same institutions rose 35%. [2]

Known limitations

BioCatch researchers compiled the figures from reports filed by institutions using the company’s fraud-detection software. [3]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   Reported victims of employment scams more than tripled over the past 12 months at more than 370 banks and other financial institutions in 21 countries.

2. [2]

   The 258% rise outran every other scam type, while total reported scams across the same institutions grew 35%.

3. [3]

   Scams by the numbers (Source: BioCatch) Researchers at BioCatch, a fraud-detection vendor, compiled the figures from reports filed by the institutions that use its software.

Read the original article →

What happened

MetaMask said it was responding to an ongoing security incident affecting part of its infrastructure. [1]

MetaMask said it was addressing and remediating the issue internally with external partners and security advisers. [2]

Why it matters

MetaMask said it had identified no immediate threat to MetaMask wallets at that time. [3]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   MetaMask on Thursday said it's responding to what it described as an "ongoing security incident" impacting part of its infrastructure.

2. [2]

   "We are actively addressing and remediating the issue internally, in coordination with external partners and security advisors," the software cryptocurrency wallet maker said.

3. [3]

   "At this time, we have identified no immediate threat to MetaMask wallets." MetaMask

Read the original article →

What happened

Bitget confirmed that attackers stole $387.5 million and said the theft involved a zero-day flaw in third-party security products, based on ongoing findings from SlowMist. [1]

The investigation identified malicious activity involving third-party security products, including a zero-day vulnerability, and recovered a customized attacker tool. [2]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   Cryptocurrency exchange Bitget on Wednesday confirmed that attackers who stole $387.5 million last week exploited a zero-day flaw in third-party security products, citing ongoing investigation findings from SlowMist.

2. [2]

   "Their investigation identified malicious activity involving third-party security products, including a zero-day vulnerability, and recovered a customized tool used by the attacker

Read the original article →

What happened

The article describes junior SOC analysts as gaining experience through repeated phishing, malware-pattern, and case-note work, while AI tools handle much of that repetitive activity. [1] [2] [3]

Repeated exposure to familiar work helps analysts recognize what normal activity looks like and notice deviations. [2] [4]

Why it matters

The article’s title states that many expect AI in the SOC to make entry-level security jobs harder to obtain. [3]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   A junior analyst in a security operations center, or SOC, has usually learned the job the slow way.

2. [2]

   You work the same phishing lure dozens of times, chase the same familiar malware pattern and write up the same case note at the end of the shift.

3. [3]

   AI tools handle a lot of that repetitive work, and the people doing the … More → The post Many expect AI in the SOC to make entry jobs harder to get appeared first on Help Net Security .

4. [4]

   Eventually you know what normal looks like, which is how you notice when something isn’t.

Read the original article →

What happened

Researchers published the first public proof-of-concept for CVE-2026-86950, an Apple CoreGraphics flaw that Apple says may have been used against specific targeted individuals. [1]

A malicious PDF containing a crafted embedded font crashes unpatched iPhones and Macs. [2]

Why it matters

The published code causes a crash rather than code execution. [3]

Known limitations

The evidence is truncated before explaining whether the memory corruption can be converted into working exploitation. [4]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   Security researchers have published the first public proof-of-concept for CVE-2026-86950, an Apple CoreGraphics flaw Apple says may have been used in attacks against specific targeted individuals.

2. [2]

   The trigger is a malicious PDF with a crafted embedded font that crashes unpatched iPhones and Macs.

3. [3]

   The code causes a crash, not an execution error.

4. [4]

   Turning the memory corruption into a working

Read the original article →

What happened

BlackFog announced ADX Vision 2.0 to help organizations govern and control generative and agentic AI use across the enterprise. [1]

The source identifies prompt injection and unauthorized data exposure among the risks that traditional controls were not designed to address; the source text is truncated after an additional risk reference. [3]

Why it matters

The source says organizations face a changing security challenge as employees deploy agents capable of making decisions and acting on corporate data. [2]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   BlackFog has announced the launch of ADX Vision 2.0, expanding its AI security capabilities to help organizations govern and control the use of generative and agentic AI across the enterprise.

2. [2]

   As employees move from simply interacting with AI tools to deploying agents capable of making decisions and acting on corporate data, the security challenge for organizations is changing.

3. [3]

   Traditional controls were not designed to address risks such as prompt injection, unauthorized data exposure and limited … More → The post BlackFog adds prompt protection and governance for agentic AI appeared first on Help Net Security .

Read the original article →

What happened

Thales announced Sentinel Envelope Plus, an addition to its Sentinel Envelope software protection solution, designed to harden compiled applications against AI-assisted reverse engineering, automated zero-day vulnerability discovery, and automated exploit generation. [1]

Sentinel Envelope Plus applies multiple protection layers to software applications without requiring source-code changes or special compilation environments. [2]

Why it matters

The source states that AI-assisted tools are reducing the specialist expertise and time needed to analyze software for vulnerabilities. [3]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   Thales has announced Sentinel Envelope Plus, a new addition to its Sentinel Envelope software protection solution that significantly hardens compiled applications against AI-assisted reverse engineering, automated zero-day vulnerability discovery, and automated exploit generation.

2. [2]

   Sentinel Envelope Plus applies multiple layers of protection to software applications, without requiring source code changes or any special compilation environments.

3. [3]

   AI-assisted tools are making it faster and easier to analyze software for vulnerabilities, reducing the specialist expertise and time required for … More → The post Sentinel Envelope Plus adds software protection without source code changes appeared first on Help Net Security .

Read the original article →

What happened

Cisco disclosed that attackers exploited CVE-2026-76504, a vulnerability in its SD-WAN solution, in zero-day attacks. [1]

Cisco incident responders became aware of active exploitation in September 2026 after a Cisco Technical Assistance Center support case was reported and resolved. [2]

Why it matters

Cisco has not shared details about the attacks, but it provided indicators of compromise for defenders to use when checking for possible impact. [3]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   For the fifth time this year, Cisco revealed attackers have exploited a vulnerability (CVE-2026-76504) in its SD-WAN solution in zero-day attacks.

2. [2]

   The vendor’s incident responders became aware of active exploitation of this vulnerability in September 2026, after getting pinged and resolving a Cisco Technical Assistance Center (TAC) support case.

3. [3]

   Cisco has yet to share any details about the attacks, but it has provided indicators of compromise defenders should look for to check whether they have … More → The post New Cisco SD-WAN zero-day exploited in-the-wild (CVE-2026-76504) appeared first on Help Net Security .

Read the original article →

What happened

The source describes a recurring security discussion in financial-services organizations about eliminating a class of vulnerabilities and upgrading the platform where they occur. [1] [2]

The source says proposed platform upgrades can prompt regression-testing costs and concerns about the change-freeze calendar. [2] [3] [4]

The source states that the finding may ultimately receive an exception, a compensating control, and a date. [5]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   Every security leader at a bank, insurer, or asset manager has had a version of this conversation: Security wants to eliminate a class of vulnerabilities.

2. [2]

   Engineering explains what it would take to upgrade the platform where they live.

3. [3]

   Somebody prices out the regression testing.

4. [4]

   Somebody else raises the change-freeze calendar.

5. [5]

   The finding gets an exception, a compensating control, and a date

Read the original article →

What happened

The Defense Manpower Data Center is notifying millions of people that hackers gained access to their personal data. [1]

The breach affects 2.76 million living individuals and 294,000 deceased individuals, according to a Defense Department official cited by CNN. [2]

Why it matters

Affected groups can include current and former defense personnel and their dependents. [2]

DMDC is a central hub for US Department of Defense data on military personnel, service status and benefits eligibility. [3]

Known limitations

The supplied evidence does not specify which personal data was accessed, how attackers obtained access, or what remediation is available. [1] [2]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   The Pentagon’s Defense Manpower Data Center (DMDC) is notifying millions of people that hackers gained access to their personal data.

2. [2]

   The breach affects 2.76 million living individuals, a group that can include current and former defense personnel and their dependents, along with 294,000 deceased individuals, a Defense Department official told CNN.

3. [3]

   Established in 1974, DMDC serves as a central hub for US Department of Defense data on military personnel, service status and benefits eligibility.

Read the original article →

What happened

Legit Security announced that its Agentic Remediation capability now covers vulnerabilities in open-source dependencies as well as first-party code. [1]

The capability is described as enabling development teams to move from vulnerability detection to a verified fix without manual triage. [1]

Why it matters

The expansion is presented as addressing application-security exposure associated with the widespread use of open-source dependencies in modern codebases. [2]

The source links each newly added package with potential exposure to known vulnerabilities, in the context of accelerated AI-generated software delivery. [2]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   Legit Security has announced an expansion of its Agentic Remediation capability to cover vulnerabilities found in open-source dependencies, not just first-party code, enabling development teams to move from vulnerability detection to a verified fix without manual triage.

2. [2]

   The expansion addresses a growing gap in application security: as AI-generated code accelerates software delivery, most modern codebases are made up largely of open-source dependencies, and every new package introduces potential exposure to known vulnerabilities.

Read the original article →

What happened

An agentic AI-powered attack against the Dutch Institute for Vulnerability Disclosure on September 21 exploited two zero-day vulnerabilities in Zammad, an open-source helpdesk and customer-support ticketing system. [1]

The two flaws reportedly enabled session hijacking, remote code execution, and privilege escalation from a Zammad user to root within seconds. [2]

Why it matters

After reaching root, the attackers were able to access other services and read additional information; the supplied excerpt truncates the remainder of this account. [3]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   An agentic AI-powered attack that hit the Dutch Institute for Vulnerability Disclosure (DIVD) on September 21 exploited two zero-day vulnerabilities in Zammad, an open-source helpdesk and customer support ticketing system.

2. [2]

   “Used together, [the two flaws] allowed the attackers to hijack sessions, run code remotely and escalate privileges from the Zammad user to root, in seconds, due to the agentic part of this hack.

3. [3]

   From there they were able to access other services and read and … More → The post AI agent used Zammad zero-days to breach Dutch vulnerability disclosure non-profit appeared first on Help Net Security .

Read the original article →

What happened

Sophos launched Sophos CISO Advantage, an agentic AI-enabled solution intended to connect security operations with security strategy. [1]

The solution provides organizations with a picture of cyber risk, a prioritized risk-reduction plan, and measurable evidence of progress in plain language for business leaders. [2]

Why it matters

Sophos describes CISO Advantage as turning security data into strategy and measurable improvement through agentic AI. [3]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   Sophos has launched Sophos CISO Advantage, an agentic AI-enabled solution that connects security operations to security strategy.

2. [2]

   The solution gives organizations a picture of their cyber risk, a prioritized plan to reduce it, and measurable proof of progress, in plain language that business leaders can understand, fund, and act on.

3. [3]

   Sophos CISO Advantage defines a new category in the market, turning security data into strategy and measurable improvement, driven by agentic AI.

Read the original article →

What happened

Google announced Gemini 4 Argon, initially rolling it out to trusted cyber defenders through the Fairwind Program rather than releasing it directly to the public. [1] [2] [3]

Argon has a 1-million-token output limit, compared with 64,000 tokens for the previous generation; Google attributes this expansion to longer, more complex reasoning trajectories. [4] [5] [6]

Reported Google and partner applications include optimizing a quantum-computing process by 40%, freeing more than 300 TiB through data-center memory optimizations, and translating C and C++ code into Rust. [7] [8] [9]

For the libgav1 decoder, Argon replaced 32,000 lines of SIMD code; the resulting Rust version reportedly runs 2.7 times faster than the previous Rust version, produces the same video output, and remains memory-safe. [10] [11]

For trusted security teams and Google engineers, Argon operates without its usual cyber safety restrictions so it can actively search for vulnerabilities for defenders to fix. [16] [17] [18]

Why it matters

The article reports that Argon found a critical flaw in healthcare software used by hospitals worldwide that could expose personal data, after earlier AI models failed to find the same issue. [12] [13] [14]

Argon scored 68% on CWE-bench v1, tying for first place on the article’s measure of AI vulnerability-fixing performance. [15]

Known limitations

Before broader rollout, Google says it is hardening defenses against cyber or CBRN misuse, indirect prompt injection, and risks in sandboxed environments used for high-risk training and testing. [19] [20]

Google says Argon’s chain-of-thought and actions are monitored in real time, with execution halted when necessary if behavior goes beyond the user’s intentions; the article also reports that captured monitoring data is excluded from training. [21] [22]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   Google unveils Gemini 4 Argon, a frontier AI model built for coding, enterprise work, and autonomous cybersecurity defense, rolling out to trusted testers.

2. [2]

   Google announced Gemini 4 Argon, and it’s not going straight to the public.

3. [3]

   It’s rolling out first to a set of trusted cyber defenders through what Google calls the Fairwind Program, which tells you something about where the company thinks this model’s sharpest edge actually is.

4. [4]

   More interesting than the price is the output limit: 1 million tokens, up from 64,000 on the previous generation.

5. [5]

   “To support Gemini 4 Argon’s capabilities across longer, more complex use cases, we are significantly expanding the model’s output token limit to an industry-leading 1M tokens, up from the previous 64K tokens.” reads the announcement .

6. [6]

   “When the model has the headroom to think deeply and generate hundreds of thousands of tokens in a single trajectory, it adds a new level of depth in reasoning to solve tough problems in one go.” Google engineers are already using Argon internally, with some impressive results.

7. [7]

   In one case, Argon helped quantum computing researchers optimize a resource-heavy process and improve the published baseline by 40% in just a few minutes.

8. [8]

   In another, Argon agents analyzed data from Google’s data centers and found memory optimizations that freed more than 300 TiB after deployment.

9. [9]

   Argon is also being used to rewrite C and C++ code in Rust.

10. [10]

    For the open-source libgav1 video decoder, Argon replaced 32,000 lines of SIMD code through repeated testing and compiler analysis.

11. [11]

    The new version runs 2.7 times faster than the previous Rust version, produces the same video output, and remains memory-safe.

12. [12]

    Wiz is already using Argon through its Scan for Good program, which looks for serious vulnerabilities in public systems and fixes them for free.

13. [13]

    Argon found a critical flaw in healthcare software used by hospitals worldwide that could expose personal data.

14. [14]

    Earlier AI models had examined the same issue but failed to find it.

15. [15]

    On CWE-bench v1, a test that measures how well AI models can fix vulnerabilities, Argon scored 68%, tying for first place.

16. [16]

    Cybersecurity is one of the main areas where Google is focusing Argon.

17. [17]

    For trusted security teams and Google’s own engineers, Argon runs without its usual cyber safety restrictions.

18. [18]

    This allows it to actively look for vulnerabilities that defenders can then fix.

19. [19]

    None of this means Google is releasing a model that can autonomously hack things and calling it a day.

20. [20]

    Before any broader rollout, the company says it’s hardening defenses on four separate fronts: blocking misuse for cyber or CBRN attacks while still allowing legitimate dual-use research, improving resistance to indirect prompt injection, where hidden instructions try to hijack the model’s behavior, and sealing off the sandboxed environments used for high-risk training and testing before anything risky happens inside them.

21. [21]

    Argon’s chain-of-thought and actions get monitored in real time, with execution halted if the model starts drifting past what the user actually asked for, and Google is careful not to feed what that monitoring catches back into training, specifically so the model doesn’t learn to reason its way around its own watchers.

22. [22]

    “In order to prevent Argon from stepping out of bounds to try to accomplish a task in a way that goes beyond the user’s intentions, we are deploying misalignment mitigations that monitor Argon’s chain-of-thought and actions and stop execution when necessary.” concludes the announcement.

Read the original article →

What happened

RadarFirst announced the general availability of Radar AI Incident Management, described as a purpose-built solution for investigating, managing, and documenting AI-related incidents. [1]

The article’s headline and text associate RadarFirst’s offering with investigating AI bias, data exposure, and unintended actions. [4]

Why it matters

The source identifies AI-related adverse events that may involve privacy, security, legal, compliance, and product teams as organizations deploy AI across multiple business contexts. [2]

Examples of incidents named by the source include harmful outputs, biased outcomes, sensitive-data exposure, unexpected AI actions, and policy failures that may require coordinated response. [3]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   RadarFirst has announced the general availability of Radar AI Incident Management, a purpose-built solution that helps organizations investigate, manage, and document AI-related incidents.

2. [2]

   As organizations deploy AI across customer experiences, employee workflows, business operations, and decision-making processes, adverse events can create risks that span privacy, security, legal, compliance, and product teams.

3. [3]

   Harmful outputs, biased outcomes, sensitive data exposure, unexpected AI actions, and policy failures can quickly require a coordinated response.

4. [4]

   Radar AI Incident Management provides … More → The post RadarFirst helps teams investigate AI bias, data exposure and unintended actions appeared first on Help Net Security .

Read the original article →

What happened

DeepKeep announced AI Lens for Developers, an extension of its AI usage-control and runtime-protection modules for software developers and coding agents that can write, modify, and execute code. [1]

The capability provides security teams with policy enforcement, audit visibility, and runtime security for coding agents including Cursor and Claude Code. [2]

The headline states that AI Lens flags coding-agent data leaks and routes destructive commands for approval. [3]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   DeepKeep has announced AI Lens for Developers, a new extension to the company’s AI usage control and runtime protection modules to secure software developers and their coding agents that can write, modify, and execute code on their behalf.

2. [2]

   The capability gives security teams policy enforcement, audit visibility, and runtime security over coding agents such as Cursor and Claude Code, closing a critical gap most security programs have ever had to cover before.

3. [3]

   90% of developers … More → The post DeepKeep’s AI Lens flags coding agent data leaks and routes destructive commands for approval appeared first on Help Net Security .

Read the original article →

What happened

Eurojust says KillSec has been active since 2024 and is responsible for almost 1,000 attacks worldwide; a 16-year-old is suspected of being its main operator. [1] [2]

The group reportedly entered organizations’ systems by exploiting poorly secured access, particularly access linked to cloud storage. [3]

After gaining access, KillSec stole data and copied it to infrastructure it controlled. [4]

Known limitations

The supplied excerpt truncates the account of what KillSec threatened to do with the stolen data and provides no further arrest or investigation details. [5]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   A 16-year-old is suspected of being the main operator of KillSec, a ransomware group that Eurojust says is responsible for almost 1,000 attacks worldwide.

2. [2]

   Seizure notice (Source: Eurojust) According to Eurojust, KillSec has been active since 2024.

3. [3]

   The group got into organizations’ systems by exploiting poorly secured access, particularly access linked to cloud storage.

4. [4]

   “Once inside, the KillSec group stole data and copied it to their own infrastructure.

5. [5]

   They then threatened to make the stolen … More → The post 16-year-old suspected leader of KillSec ransomware group arrested appeared first on Help Net Security .

Read the original article →

What happened

Proofpoint attributed phishing campaigns to TA419, a China-aligned cyber espionage group, targeting U.S. AI policy experts and seeking access to cloud accounts at think tanks, universities, and law firms. [1] [2] [3]

The campaigns impersonated prominent officials, economists, and an Anthropic employee, using invitations about AI policy committees, export controls, supply chains, or military use of Claude models to start conversations. [4] [5] [6] [7] [8]

After a target replied, shortened links redirected through multiple websites to a false Microsoft OneDrive sign-in page intended to capture credentials and active browser sessions. [9] [10] [11] [12]

Proofpoint said TA419 has targeted people connected to U.S. and Japanese think tanks, defense contractors, universities, and law firms since at least April 2025, and registered domains resembling real organizations. [19] [20] [21]

Why it matters

Proofpoint characterized the operation as adversary-in-the-middle phishing: victims could interact with genuine Microsoft infrastructure, enter passwords, complete multifactor authentication, and pass access checks while session information was captured. [13] [14] [15]

TA419 used a modified Frameless BitB tool to display a fake Microsoft login window over a page resembling a OneDrive document-sharing site. [16] [17] [18]

Known limitations

The report did not identify victims or state whether any accounts were compromised, and it did not directly link the activity to the Chinese government. [22] [23]

Cited source evidence

These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.

1. [1]

   A China-aligned cyber espionage group targeted U.S.

2. [2]

   artificial intelligence policy experts through phishing emails that impersonated prominent officials, economists and an employee of AI company Anthropic, according to research released Thursday by Proofpoint.

3. [3]

   The campaigns, which the cybersecurity company attributed to a group it calls TA419, sought access to cloud accounts held by people at think tanks, universities and law firms.

4. [4]

   Proofpoint said the group began a campaign in July by impersonating Lynne Parker , a former principal deputy director of the White House Office of Science and Technology Policy, and economist and foreign policy expert Heidi Crebo-Rediker.

5. [5]

   The emails invited recipients to join a supposed AI policy advisory committee or contribute to a report on AI export controls and supply chains.

6. [6]

   Proofpoint also identified a February campaign in which the same group impersonated a senior Anthropic employee.

7. [7]

   That message asked an AI policy analyst at a U.S.

8. [8]

   think tank for feedback on the military’s use of Anthropic ’s Claude AI models, which was a highly controversial topic at time.

9. [9]

   The initial messages did not immediately request passwords or direct recipients to a sign-in page.

10. [10]

    After a target replied, the group sent a shortened link said to contain more information.

11. [11]

    The link redirected recipients through several websites before leading to a false Microsoft OneDrive sign-in page.

12. [12]

    Proofpoint said the setup was intended to capture account credentials and active browser sessions.

13. [13]

    The firm described the operation as an adversary-in-the-middle phishing attack.

14. [14]

    In such attacks, the victim interacts with genuine Microsoft infrastructure during part of the process, looking and behaving like a legitimate sign-in.

15. [15]

    The person may enter a password, complete a multifactor authentication prompt and pass access checks while the attacker captures the session information created by the login.

16. [16]

    Proofpoint said TA419 used a modified version of an open-source phishing tool known as Frameless BitB .

17. [17]

    The tool creates a false browser window within a webpage, imitating a familiar sign-in prompt.

18. [18]

    In this case, it was used to present a fake Microsoft login window over a page that resembled a OneDrive document-sharing site.

19. [19]

    Proofpoint said TA419 has targeted individuals connected to U.S.

20. [20]

    and Japanese think tanks, defense contractors, universities and law firms since at least April 2025.

21. [21]

    The group also registered domains resembling real organizations, including the Heritage Foundation, the World Economic Forum and the Japan-Taiwan Exchange Association.

22. [22]

    The report did not identify victims or state whether any accounts were compromised.

23. [23]

    The report does not directly link the activity to the Chinese government.

Read the original article →