As security operations teams now use large language models (LLMs) and autonomous AI agents into their daily work, a new frontier is emerging: attackers deliberately manipulating AI agents. Prompt injection attacks—where an attacker hides malicious instructions that cause an AI agent to ignore its safety rules—pose a serious risk to enterprises. These attacks continue to grow in size and scale.  

Snyk’s security audit of the Agent Skills ecosystem, which includes Anthropic’s Claude, Vercel, and others, that 36% of all skills contained at least one critical-level security issue, including malware distribution, prompt injection attacks, and exposed secrets.

In June, researchers at Mozilla tested a prompt injection attack on Claude using indirect prompt injection—a technique that embeds malicious instructions in external content the AI agent processes. In this proof-of-concept, attackers took over developers’ systems by hiding indirect prompts in normal-looking repositories. When Claude Code executed them, the agent spawned a reverse shell.

AI agents often connect to more sensitive data than human employees do., A successful prompt injection can lead to catastrophic data loss or unauthorized system actions. Defending against prompt injection attacks requires multiple layers of protection. Security teams must monitor agent behavior for anomalies and prepare for agent containment, forensic preservation, and system remediation. Because AI agents execute tasks at machine speed, human responses must be able to match that pace.

The architecture of trust: Protocols and no “black box”

AI-native workflows need governed access rather than “black-box” autonomy. Modern governance frameworks use standardized protocols like the Model Context Protocol (MCP) to provide secure communication between AI clients and data sources. Visibility and transparency in agentic AI workflows matter, especially in cybersecurity. Autonomous agents perform complex tool executions and use independent logic, so they must show how they reached their decisions to meet regulatory requirements. Agents without transparency post serious risks: obscured reasoning can trigger unpredictable tool interactions, bypass governance controls, and create uncontrolled defensive gaps.

Implementing these protocols matters:

• Bounded Tenant Awareness: In a stable agentic AI architecture, multi-tenancy scales well. But if an AI tenant misbehaves, the entire system can fail. Bounded tenant awareness isolates any misbehaving AI agent to prevent cross-tenant contamination or data leakage.
• Strict Access Controls: By controlling connections to the platform, organizations can stop “ignore previous instructions” style bypasses. Maintain tight control over what the AI can see and do within a workflow.
• Standardized Telemetry: All telemetry must remain consistent and audit-ready. Even if an AI interaction is attempts to break rules, the underlying data movement gets tracked against established frameworks like MITRE ATT&CK and NIST.

Detecting the aftermath: UEBA and NDR as safeguards

A robust, unified SecOps platform can detect anomalous behavior even after prompt injection tricks an AI agent. Prompt injections often serve to steal credentials theft or extract data. When detected it’s important to act quickly. In agentic AI systems, misbehavior can escalate privileges, manipulate memory layers, create unauthorized identities, or alter shared reasoning components. Containment must be automatic and enforced at identity, authentication, and authorization layers.

These safeguards include:

• User and Entity Behavioral Analytics (UEBA): Identity-focused correlation and behavioral baselines to identify anomalous user activity or privilege escalation. If a compromised AI agent acts outside of its normal operational parameters, UEBA flags it in real-time and alerts a human security analyst.
• Network Detection and Response (NDR): Combining network traffic analytics with endpoint and cloud telemetry, NDR can identify data exfiltration or policy violations from a successful prompt injection.
• Multi-Layer AI Filtering: AI filters reduce raw alerts into high-fidelity incidents, cutting noise by up to 90%. This keeps the signals of an AI-driven attack from disappearing in a busy SOC.

Humans remain the strongest defense against AI agent social engineering. The human security analyst is still the one who makes the final decision. While AI handles triage and correlation, humans retain final control over response actions.

Moving beyond reactive guardrails

The traditional SOC model was never designed to handle machine-speed, AI-driven attacks. A human-augmented autonomous SOC approach moves from reactive alert handling to a proactive, verdict-first model. By combining a transparent, governed AI access with robust UEBA and NDR, organizations keep the SOC secure, transparent, and resilient as social engineering methods target machines.

The post Why transparent AI agents matter more than you think appeared first on CyberScoop.

Alexander Martin reports: A 20-year-old man was sentenced Monday to two years in prison for coercing more than 100 girls worldwide into sexual and self-harm activity using platforms such as Snapchat, Telegram and Discord, British authorities said. Justin Swaddle, of Leeds in northern England, targeted 117 female victims aged 13 to 17, according to the...

Remote, unauthenticated attackers could exploit the bugs to cause a denial-of-service (DoS) condition.

The post Cisco Warns of High-Severity ClamAV Vulnerabilities With Public PoC appeared first on SecurityWeek.

CERT.PL said this appears to be the first instance of a private APN being used as an attack vector.

The post Novel Private APN Pivot Let Hackers Sabotage Second Polish Energy Facility appeared first on SecurityWeek.

Researchers believe they've found evidence that brain signals reveal the pitches of notes in imagined musical melodies, reports Medical Express: Study participants were 10 epilepsy patients with electrodes on the surfaces of their brains for clinical monitoring. Participants quietly listened to the first bars of children's tunes, then imagined the melodies of subsequent bars before humming their imagined melodies. Brain activity from the patients carried information about relative pitches within the imagined melodies. Combining note-level pitch predictions over time allowed the research team to recreate melodic changes that reflected the structure of the imagined songs... "The model we created decodes relative pitch classes — such as do, re, mi, fa, sol, and la — rather than exact, absolute pitches," [said tudy co-lead Jii Kwon of Seoul National University]. "This is useful because people often recognize melodies by the relationships between notes, even when the same melody is played in a different key."

Read more of this story at Slashdot.

As AI-generated code continues to be injected into all corners of the internet, concerns have risen about an expanding attack surface for malicious hackers to exploit.

Some have argued that the enhanced cybersecurity capabilities of large language models could serve as a check, finding and fixing vulnerabilities nearly as fast as they’re created.

But new research that tested the patching capabilities of two popular commercial models, OpenAI’s ChatGPT 5.5 and Anthropic’s Claude Opus 4.8, found that generative AI is more likely to create an exploitable patch or introduce entirely new bugs than close off a vulnerability.

Researchers at 1Password tested the models ability to patch six “high-impact, high-complexity” CVEs, including the “Copy Fail” vulnerability, a kernel flaw that can give an attacker root access to Linux cloud environments. The overall success rate (or fully patching the vulnerability without introducing new problems), was less than a coin flip at 47%.

“Our research findings show that, in aggregate across a variety of scenarios, both Claude and ChatGPT had a low rate of successful patch generation, which we define as full remediation of all known exploit paths with no erroneous changes to application behavior,” wrote Keith Hoodlet, Axel Mierczuk and Spencer Michaels.

“The models often addressed only a subset of vulnerable code paths, added fragile guard code that satisfied tests while failing to address the vulnerability’s root cause, and sometimes introduced subtle changes in the application’s behavior while patching the immediate vulnerability,” the authors continued.

The research suggests that largely autonomous vulnerability-discovery and patching may not yet be effective in fixing the explosion of vulnerable code that is being created in the AI era.

Other private sector research has pointed to a similar problem. A report this year from Veracode found that while LLMs have made “enormous strides” in crafting workable code, “security is a different story.” Testing across a range of frontier models found the average security “pass rate” for AI generated code is around 56%. Newer models like GPT 5.5 push closer to 70%, while more than half sit between 50-53%.

Veracode tested 100 different models and while there was variability, in general a small number of models were showing progress on security patching while the rest have experienced “stagnation.” Similar to the 1Password research, in 44% of Veracode tests the models introduced a detectable OWASP Top 10 vulnerability into the codebase.

An important caveat: neither report tested newer models, like Anthropic’s Mythos or OpenAI’s GPT-5.6-Sol, that frontier companies tout as having significantly higher cybersecurity capabilities.

Those advanced models can identify and fix vulnerable code. Anthropic and OpenAI are distributing them to key industries through Project Glasswing and Daybreak before foreign or open-source alternatives can compete.

Tim Jarret, vice president of product at Veracode, told CyberScoop that AI tools are still subject to a range of limitations that can make them unreliable for cybersecurity patching without knowledgeable humans in the loop.

While some vulnerabilities – like SQL injections – can be easily patched through automation, other bugs like cross-site scripting, can be exploitable in several different ways and require either a human touch, additional context or both to fully close off. Additionally, models can slowly lose context from prior sessions over time, affecting their ability to complete tasks correctly and raising the possibility they’ll hallucinate to fill in the missing gaps.

“I think we would say, at this point, that Iits premature to treat those as anything other than another code change to the code base that needs to be reviewed and accepted by the team, as opposed to letting the agent merge the code freely,” said Jarrett.

However, he acknowledged that may not be possible in a world where AI agents are generating exponentially more code for human defenders to review. Some kind of automated code review will be necessary – preferably not by the same automation tool that produced the code. The ultimate goal is the same as it has always been in security: “trust but verify.”

“Ninety percent of the time, the human check might just be ‘did the cross check look good?’ Do we have a thumbs up?’” Jarrett said. “In those cases where there’s still something wrong, that’s where you focus your attention a little bit more.”

The post More than half of AI-generated patches are broken appeared first on CyberScoop.

Suno says it will begin watermarking and fingerprinting AI-generated songs, tighten download policies to curb mass distribution, and explicitly prohibit deceptive audio and unauthorized voice or likeness cloning. The changes come as the company faces mounting copyright lawsuits, a recent adverse ruling in Germany, and scrutiny over a past data breach that revealed training-data sources. TechCrunch reports: In a blog post, co-founder and CEO Mikey Shulman shared core principles and said that the platform wants to promote original creation while enabling more people to make music with its AI tools. One of the key points of contention involved users uploading AI-generated songs on other streaming platforms and gaming the system to earn revenue. Suno said that now it will use audio watermarking and fingerprinting to prevent misuse on other streaming platforms. It's not clear if Suno will use an existing system like Google's Synth ID or adopt a new one, and the company did not say when contacted by TechCrunch about this. The startup also signed an agreement with lyrics provider Musixmatch to use its Sentinel system for copyright detection, the blog post said. [...] The company added that it plans to add a new download policy to bar mass distribution on streaming platforms, but declined to provide details on the record. Suno has also changed its community guidelines to explicitly prohibit "deceptive audio presented as real" and "using a real person's voice or likeness without permission" to prevent copycats.

Read more of this story at Slashdot.

(Video) In this podcast, we share insights from Edna Conway, a recognized leader in cybersecurity and supply chain resilience with over 40 years of experience in the field.

The post Podcast: Compliance Won’t Save You: The Future of Cyber Risk with Edna Conway appeared first on SecurityWeek.

Patches were rolled out for two dozen vulnerabilities, including one with public proof-of-concept (PoC) code.

The post Cisco Patches Critical SD-WAN, IOS XE, FMC Vulnerabilities appeared first on SecurityWeek.

AI Security Institute reports Anthropic and OpenAI models going rogue against real people, organizations, and open source projects.

The post AI Agents Targeted Real People and Projects During Cybersecurity Tests appeared first on SecurityWeek.

The UK’s AI Security Institute said that their AI research system took “unsanctioned” actions over the internet to engage in “sustained, potentially harmful activity directed at real people.”

According to a blog post and technical report released Tuesday, the organization noticed “unusual data transfers” leaving their research systems through the pseudo-anonymous Tor network on July 28. The models, which included Anthropic’s Mythos 5 and OpenAI’s GPT-5.6-Sol models, were being tested for their cybersecurity capabilities.

Like the OpenAI and Anthropic incidents before it, the offending agents were attempting to solve a challenge. On 10 of 122 runs taken by two models, they took a combined 19 malicious actions, including attempting to insert malicious code into a real open-source software project and creating fake online identities to contact human maintainers and press them to approve it.

It also inserted malicious prompt injection instructions in places “where it reasoned that other automated AI systems might pick them up and execute them.” Sometimes multiple agents would collaborate on a task, with one model leaving public messages or notes for the other on GitHub.

However, AISI said their incident differs in one critical way.

“Importantly, this was not a case of a model escaping its secure test environment, or ‘sandbox,’” the institute wrote in an unsigned blog. “As was standard in our cyber testing, we had intentionally permitted internet access, and model-provider cyber classifiers were deliberately disabled – conditions that do not reflect how frontier models are made available to the public.”

The institute said that while their own design and evaluation choices “to some degree” enabled the malicious behavior, the models still displayed “novel, potentially deceptive behaviors” and were executed “to an extent and severity that we did not anticipate.”

The same day, OpenAI acknowledged AISI and another third-party cybersecurity tester,  private firm Irregular, had recently reported incidents where their models exceeded “beyond their intended testing boundaries.”

That included GPT-5.6-Sol reusing GitHub tokens left by another agent to attempt to recover accounts and to access a DNS server that contained malicious payloads meant to exploit a software vulnerability.

The setup did not work, there is no evidence any real resolver queried it, and the infrastructure was removed when the evaluation ended,” the company wrote in an unsigned blog post Tuesday.

OpenAI said it would review its own third-party testing procedures to focus on higher risk evaluations, assess requests by third-parties to enable internet access, stop conditions and other features for their models.

The incident with Irregular occurred on July 29 during a Capture-the-Flag cybersecurity evaluation of OpenAI’s models. Due to a “misconfiguration” the models were allowed to access the public internet, where they encountered and exploited a real domain, mistakenly believing it was still in a test environment.

OpenAI said an investigation by Irregular is ongoing, but also found that the models had found and used credentials for the site at one point. The blog also references other additional potential cybersecurity incidents.

“Irregular has informed us that all of the issues identified pertaining to the incident are no longer active and relevant safeguards were added to the testing environment,” the blog said. “Irregular has also communicated about related incidents involving other labs from the same testing environment.”

CyberScoop has reached out to Irregular for comment.

The incidents were made public the same day that the White House met with Anthropic, Open AI and other frontier AI companies to preview a new framework for evaluating models before they’re released publicly. Some media outlets have reported that after an executive order, export controls and other actions, the administration does not plan to make the new framework public.

The post AISI, OpenAI report more ‘unsanctioned’ model hacks appeared first on CyberScoop.

The Trump administration’s haphazard and opaque interventions into artificial intelligence security matters could catapult Chinese alternatives into broader acceptance, posing new security risks altogether, a group of Democratic senators wrote to top administration officials Monday.

The five senators said that the administration’s handling has alternated between too passive, such as when OpenAI models escaped testing in the Hugging Face hack last month, and overstepping, such as when the Commerce Department suspended access for any foreign national to Anthropic’s Fable 5 and Mythos 5 in June.

“The Administration’s ad hoc and unpredictable approach undermines U.S. competitiveness, heightening market incentives to adopt open weight models from vendors based in the People’s Republic of China (PRC),” wrote Sens. Kristen Gillibrand of New York, Adam Schiff of California, Mark Warner of Virginia, Chris Coons of Delaware and Mark Kelly of Arizona.

In the Hugging Face hack, the senators wrote that “the Federal Government cannot be passive as these capabilities emerge.”

In the case of the Fable 5 and Mythos 5 suspensions, the senators said that the administration “utilized an infrequently used authority to direct Anthropic to suspend all access to its Fable 5 and Mythos 5 models for foreign nationals (including foreign national employees inside the United States) citing an undisclosed national security concern later described as a narrow jailbreak finding.”

Because Anthropic couldn’t immediately assess users’ nationality, the firm had to disable both models for everyone. The administration and Anthropic negotiated for 18 days behind closed doors before reaching an agreement, the lawmakers complained.

“While the Administration may have been responding to real security concerns to protect the United States, even justifiable interventions can create broader harm if the standards and decision-making processes are opaque, ad hoc, or unpredictable,” they said in their letter to leaders in the White House, Office of the National Cyber Director and departments of State, Treasury and Commerce. “Moreover, when the Executive Branch exercises authority delegated from Congress, such as in the conduct of export control administration, it is essential that it keep Congress fully apprised of its actions and procedures.”

During the time Anthropic was under export controls, the stock price of “an entity-listed Chinese lab” nearly doubled, the senators said. And while Hugging Face was breached, the company “had to” rely on a Chinese open-weight model due to guardrails on U.S. frontier models.

“If American models are perceived as subject to sudden access disruptions based on a black-box U.S. Government process, or as unreliable because U.S. AI labs are overcorrecting in the face of this black-box process, companies and governments in the United States and abroad may hedge by adopting Chinese or other foreign models instead,” the senators contended. “That outcome would undermine U.S. technological leadership while increasing exposure to systems that may carry risks of PRC or otherwise directed censorship, espionage, IP theft, and other supply chain security risks.”

Their letter asked for answers to questions about the standards the administration uses to determine the national security risks a frontier model presents, what legal authorities it will use to invoke restrictions, which agencies are responsible for which decisions and more.

None of the offices or departments the letter was addressed to immediately responded to a request for comment.

The letter follows inquiries at the state level, where 15 attorneys general asked OpenAI for more information regarding the security incident at Hugging Face.

The post Dem senators criticize Trump administration decisionmaking on AI security risks appeared first on CyberScoop.

Over 24,000 internet-accessible server-management interfaces disclose authentication hashes before login.

The post Decades-Old BMC Vulnerability Exposes Thousands of Data Centers to Attacks appeared first on SecurityWeek.

Charmian Aw, Ciara O’Leary, and Florence Seow of Hogan Lovells Cadwalader write: Singapore’s Personal Data Protection Commission (“PDPC”) has issued its final Advisory Guidelines on Use of Personal Data in Generative AI (“Guidelines”), marking a significant development in Singapore’s AI governance framework. The Guidelines provide long-awaited clarity on how the Personal Data Protection Act 2012...

President Donald Trump blamed Minnesota Friday for the cyberattacks its water systems have suffered in recent days, saying the state was “behind it.”

Trump said the state being “incompetent” was the issue, but it wasn’t clear whom he thought actually conducted cyberattacks that U.S. investigators have attributed to Iran — if, perhaps, somehow Minnesota incompetently cyberattacked itself. The White House referred a request for clarification back to Trump’s remarks.

“I think that Minnesota is behind it,” Trump told reporters Friday. “Because they’re grossly incompetent. I don’t think there was an Iranian cyberattack. I think Minnesota ought to get its act together.”

The White House also didn’t clarify whom the president believed was behind similar attacks in other states, when asked for comment. Trump has repeatedly used federal power aggressively in Minnesota, a state led by Gov. Tim Walz, a Democrat who was on the ticket that ran against him in 2024 as the vice presidential nominee. Trump also has downplayed Iranian attacks amid the war he launched against the nation with Israel in February.

A number of cyber experts quickly pushed back on Trump’s comments after he made them.

“Victim blaming in cyber is so 2000 and late,” cybersecurity pioneer Chris Wysopal, Veracode co-founder and chief security evangelist, said on the Bluesky social media platform. Said Jake Williams, a member of the IANS faculty: “His own intelligence services are attributing this to Iran.”

Andy Jabbour — founder and CEO of Gate 15, a cybersecurity firm which provides support to the water sector — told CyberScoop that, “speaking candidly, I’m not even sure what he was actually saying or suggesting Minnesota’s government did or didn’t do.”

“Attribution is tricky business,” he continued, referencing recent alerts from the Cybersecurity and Infrastructure Security Agency and others. “But logically, given an ongoing war with Iran, recent statements made by Iran-aligned threat groups, with assessments that the recent activity is aligned with recent CISA warnings, given yesterday’s statements from CISA and the FBI, random unsubstantiated allegations aimed at political opponents seem reckless and are a disservice to the American people.”

Walz struck back at Trump in a Facebook post, noting steps from his Department of Government Efficiency to slash federal funding. CISA has shrunken considerably under Trump, and his administration has pushed states to defend against cyberattacks that feds once countered.

“Trump knows exactly who is responsible for this attack, and knows that other states were hit too,” Walz said. “This is what modern warfare looks like, and it further illustrates there’s no plan to win a war with Iran.”

“DOGE took an axe to CISA and left the U.S. exposed to cyber attacks,” he continued. “Thankfully, our experts in Minnesota were able to identify the vulnerability quickly and work with local communities to stop it.”

A spokesperson for Minnesota IT Services, a state agency that has been responding to the water cyberattacks, declined to address Trump’s remarks.

“We remain focused on supporting affected communities, securing critical infrastructure and coordinating with local partners and federal officials as the investigation continues,” the spokesperson, Emily Zimmer, told CyberScoop. “We will not comment on political statements or speculate about attribution.”

Other cyber professionals declined to comment directly on Trump’s remarks, but offered thoughts on who was behind the attacks and their motives.

Bryson Bort, CEO and founder of Scythe said the evidence supports the attribution with Iran, and that it looks like hackers there found something they could exploit on the internet and seized the chance.

“This was a target of opportunity,” said Bort, co-founder of the ICS Village, a non-profit advancing awareness of industrial control system security; such systems are common in the water sector. “It wasn’t that Minnesota did something as a state to raise Iran’s ire.”

Cynthia Kaiser, a former top FBI cyber official, said that when the bureau conducts attributions, it looks at technical indicators but also who has the capability, who has conducted similar attacks in the past and what the purpose of the attacks is.

“Iran ticks all these kinds of things,” Kaiser, now senior vice president at cybersecurity firm Halcyon, told CyberScoop. “My view is, if it walks like a duck, if it talks like a duck, I strongly suspect it’s a duck. I’d be shocked if we found out it wasn’t Iran.”

Just last week, CISA updated an advisory about how Iranian hackers were targeting programmable logic controllers in the water sector and other sectors, a warning that the water industry’s information sharing and analysis center said it believed.

“WaterISAC is confident in our government partners’ assessment that the confirmed activity is aligned with the joint Cybersecurity Advisory (CSA) AA26-097A ‘Iranian-Affiliated Cyber Actors Exploit Programmable Logic Controllers Across U.S. Critical Infrastructure’ published and recently updated by CISA,” Tom Dobbins, executive director, WaterISAC, told CyberScoop. “We have evidence of earlier attacks from Iran even before this current conflict. Cyber attacks are the most viable way that Iran can directly attack our homeland, and it is logical that they would do so, especially given the challenges of absolute attribution.”

The water sector is often viewed as one of the most vulnerable critical infrastructure sectors, and Dobbins called on Congress to provide funding to provide funding for the ISAC.

Sen. Tina Smith, D-Minn., also took issue with Trump’s comments.

“The President provided an unserious response that is beneath the dignity of the office he holds. Iran’s purported cyberattack on Minnesota’s water infrastructure must be taken as a serious threat to our national security.  Smith said in a statement, adding that she’s been in touch with CISA and the FBI and was grateful to Minnesota’s IT experts. “The entire situation serves as a stark reminder of the danger this war puts us in the longer it drags on.”

Fellow Minnesota Democratic Sen. Amy Klobuchar had earlier been in touch with Sean Cairncross, the national cyber director and a Minnesota native, about the incident.

Trump has previously displayed a laissez-faire view toward other cyberattacks on the United States, such as when he’s been asked about Chinese and Russian cyberattacks and Trump shrugs them off as something America does, too.

He also has cast doubt before on his government officials’ assessments of who’s responsible for cyberattacks on the United States, such as when he asserted China rather than Russia was behind the landmark SolarWinds breach.

Updated 8/3/2026: with comments from Minnesota’s senators.

The post Trump blames Minnesota for cyberattacks on water sector, drawing pushback from cyber world appeared first on CyberScoop.

A security company’s systems were hacked after it installed a malicious Python package deployed by Claude. 

The post Prompted by OpenAI Disclosure, Anthropic Finds Its Own Models Hacked 3 Organizations appeared first on SecurityWeek.

Anthropic said it found three instances in which its models gained access to the live computer systems of outside organizations, according to a company blog post published Thursday.

The company said it began the review after OpenAI disclosed earlier this month that some of its models had exploited an unknown software flaw to escape an isolated test setup and reach production systems at Hugging Face, a platform for AI models and datasets. Anthropic checked over 141,000 evaluation runs for signs that Claude had reached the internet from environments meant to be closed off. It found six runs across three incidents, all tied to one outside testing partner, Irregular.

“Ultimately, many factors contributed to these incidents, but, consistent with a blameless postmortem culture, we’re approaching the fixes as if the responsibility were ours alone,” the blog post reads. “This begins with ensuring every part of our evaluation pipeline is secure, including the manner in which we integrate with external partners. Moving forward, it will include expanding our continuous monitoring of evaluation transcripts for unexpected behavior, improving our investigation tooling, and conducting more rigorous assurance work with the vendors we rely on.”

In each case, Claude was working on a “capture the flag” puzzle, a common way to test a model’s hacking skill. The model is told a piece of secret data sits on another machine and must be found. Anthropic‘s prompts told Claude it had no internet access. That was false. A setup error at the partner’s end left the machines connected to the open internet, and Claude found real systems while searching for its target.

Believing the real systems were part of the simulation, Claude broke into them using simple methods such as guessing weak passwords and finding unprotected access points. In the most serious case, the model extracted login credentials and reached a database holding several hundred rows of live data. 

In a second case, Claude built and uploaded a malicious software package to PyPI, a public directory for Python code, after noticing a fake company’s setup guide told new hires to install a package that did not exist. The package stayed online for about an hour and was installed on 15 systems, including one belonging to a security firm whose scanner ran the code and had its credentials stolen. 

In the third case, Claude scanned about 9,000 targets before breaking into one company’s application through exposed credentials and a database attack method known as SQL injection.

The three incidents involved different Claude models: Opus 4.7, Mythos 5, and an unreleased internal test model. Anthropic said the models behaved differently once they encountered signs their targets were real. Opus 4.7 recognized the systems were live but kept attacking anyway. Mythos 5 also noticed signs it was on the open internet but talked itself out of that conclusion, pointing to a system date reading 2026 as proof it remained in a simulation. The internal test model, the newest of the three, stopped once it concluded the target was real.

Anthropic said it began its transcript review last week, halted all cybersecurity evaluations, identified the three incidents, and notified the partner and affected organizations earlier this week. Two of the three organizations had not detected the intrusions before being contacted, the company said; outreach to the third was ongoing. 

Anthropic framed the incident as a failure of its testing setup and oversight rather than a case of a model acting on its own, noting that in each instance the model was operating as intended. 

The company said it is working with the outside evaluator METR on an independent review and plans to release a redacted transcript of the PyPI incident within a week. It also said it would tighten monitoring of test environments run by outside partners and expand review of evaluation logs, framing the changes as part of what it called a blameless review of its own processes.

“These facts give us cautious optimism that with tighter monitoring and controls around evaluation infrastructure, as well as continued investment in alignment, this type of risk can be overcome,” the blog post reads. 

The post Anthropic says its AI accidentally hacked three companies during safety tests appeared first on CyberScoop.

CISA is urging water and wastewater utilities to lock down internet-exposed controllers, days after intrusions hit dozens of Minnesota systems.

The post CISA Urges Water Sector to Protect OT After Coordinated Attacks on PLCs appeared first on SecurityWeek.

The vulnerability tracked as CVE-2026-20316 can be exploited by a remote, unauthenticated attacker to log into affected devices. 

The post Cisco Secure FMC Zero-Day Exploited in the Wild appeared first on SecurityWeek.

The startup will use the investment to expand its customer support, sales, and R&D teams.

The post Mate Security Raises $35 Million for Agentic SOC appeared first on SecurityWeek.