Ticker feed
Two critical vulnerabilities in Cursor IDE — the AI coding tool used by over half of Fortune 500 companies — could give attackers full remote code execution without any user interaction. Discovered by Cato AI Labs and dubbed "DuneSlide," both flaws carry a 9.8 CVSS score (CVE-2026-50548 and CVE-2026-50549).
The attack works through prompt injection: a victim simply types a normal prompt that accidentally pulls in attacker-controlled content — from a poisoned web search or rogue MCP server. From there, attackers can overwrite core sandbox binaries and compromise both the local machine and connected SaaS workspaces.
Cursor 3.0, out since April 2, fixes both. Every earlier version is affected, and there's no sign of exploitation in the wild. Getting there took some pushing: Cursor rejected Cato's initial report in February, saying its threat model didn't account for MCP server misuse, and only reopened the cases after the researchers escalated.
Cato says more disclosures are coming across other AI coding agents.
Source: Cybersecurity News
Two critical vulnerabilities in Cursor IDE — the AI coding tool used by over half of Fortune 500 companies — could give attackers full remote code execution without any user interaction. Discovered by Cato AI Labs and dubbed "DuneSlide," both flaws carry a 9.8 CVSS score (CVE-2026-50548 and CVE-2026-50549).
The attack works through prompt injection: a victim simply types a normal prompt that accidentally pulls in attacker-controlled content — from a poisoned web search or rogue MCP server. From there, attackers can overwrite core sandbox binaries and compromise both the local machine and connected SaaS workspaces.
Cursor 3.0, out since April 2, fixes both. Every earlier version is affected, and there's no sign of exploitation in the wild. Getting there took some pushing: Cursor rejected Cato's initial report in February, saying its threat model didn't account for MCP server misuse, and only reopened the cases after the researchers escalated.
Cato says more disclosures are coming across other AI coding agents.
Source: Cybersecurity News
Security firm Adversa AI has found a structural flaw — dubbed GuardFall — affecting 10 of 11 popular open source AI coding agents, including Hermes, OpenCode and Roo-code. Lead researcher Omer Ben Simon showed that attackers can embed old-school Bash tricks like quote removal and $IFS spacing into content agents ingest, such as a poisoned README or Makefile.
Once inside, those commands can silently steal AWS credentials or wipe dev environments, running with the developer's full account authority. The worst case is a CI pipeline with auto-execute switched on, where nobody is watching the terminal at all.
Only one agent, Continue, successfully blocked all test bypasses. Adversa recommends maintainers adopt a tokenize-and-canonicalize evaluator guard — the approach Continue uses — rather than relying on pattern-based text matching that Bash simply rewrites around. No fixes have been announced yet, so if your agents run unattended in CI, turning off auto-execute is the short-term answer.
Source: SecurityWeek
Security firm Adversa AI has found a structural flaw — dubbed GuardFall — affecting 10 of 11 popular open source AI coding agents, including Hermes, OpenCode and Roo-code. Lead researcher Omer Ben Simon showed that attackers can embed old-school Bash tricks like quote removal and $IFS spacing into content agents ingest, such as a poisoned README or Makefile.
Once inside, those commands can silently steal AWS credentials or wipe dev environments, running with the developer's full account authority. The worst case is a CI pipeline with auto-execute switched on, where nobody is watching the terminal at all.
Only one agent, Continue, successfully blocked all test bypasses. Adversa recommends maintainers adopt a tokenize-and-canonicalize evaluator guard — the approach Continue uses — rather than relying on pattern-based text matching that Bash simply rewrites around. No fixes have been announced yet, so if your agents run unattended in CI, turning off auto-execute is the short-term answer.
Source: SecurityWeek
AWS has patched a high-severity bug in the Amazon Q Developer extension that could let attackers steal cloud credentials just by getting a developer to open a malicious repository. Tracked as CVE-2026-12957 and discovered by Wiz Research, the flaw allowed Amazon Q to automatically load and execute MCP server configurations without user approval — silently, before any code review happened.
Because spawned processes inherited the developer's full environment, attackers could grab AWS credentials, API keys, SSH secrets and other sensitive environment variables. Wiz built a working proof of concept; no widespread attacks have been reported.
The fix landed in AWS Language Server version 1.65.0, which covers Amazon Q in VS Code and the other IDEs it supports. Update, then check your existing MCP configurations — patching stops new configs from auto-executing, but won't remove one a poisoned repo already dropped. Similar MCP-related vulnerabilities have also been found in Claude Code, Cursor and Windsurf.
Source: Dark Reading
AWS has patched a high-severity bug in the Amazon Q Developer extension that could let attackers steal cloud credentials just by getting a developer to open a malicious repository. Tracked as CVE-2026-12957 and discovered by Wiz Research, the flaw allowed Amazon Q to automatically load and execute MCP server configurations without user approval — silently, before any code review happened.
Because spawned processes inherited the developer's full environment, attackers could grab AWS credentials, API keys, SSH secrets and other sensitive environment variables. Wiz built a working proof of concept; no widespread attacks have been reported.
The fix landed in AWS Language Server version 1.65.0, which covers Amazon Q in VS Code and the other IDEs it supports. Update, then check your existing MCP configurations — patching stops new configs from auto-executing, but won't remove one a poisoned repo already dropped. Similar MCP-related vulnerabilities have also been found in Claude Code, Cursor and Windsurf.
Source: Dark Reading
The hacking group ShinyHunters has claimed responsibility for a major cyber-attack on the University of Nottingham, confirmed on June 10. Around 455,000 unique email addresses were compromised — the number of people affected is lower, since many had both a university and a personal address — and one expert put the haul at roughly 40 gigabytes taken from the university's student record system.
Stolen data includes passport numbers, national insurance numbers, dates of birth, financial details, ethnicity and disability information, affecting both current students and alumni. Early speculation pointed to voice phishing or a supply chain breach; reporting since has identified the entry point as an Oracle PeopleSoft zero-day, CVE-2026-35273, which Mandiant links to attacks on more than 100 organisations, most of them universities.
Security researcher Troy Hunt believes the university was held to ransom and refused to pay, which is why the data was published — the university itself won't comment while the criminal investigation runs. EMSOU's Regional Cyber Crime Unit is investigating, and the university has notified the ICO, NCSC, Office for Students and Action Fraud.
Affected individuals should enable multi-factor authentication and stay alert to unexpected phone calls.
Source: BBC News
The hacking group ShinyHunters has claimed responsibility for a major cyber-attack on the University of Nottingham, confirmed on June 10. Around 455,000 unique email addresses were compromised — the number of people affected is lower, since many had both a university and a personal address — and one expert put the haul at roughly 40 gigabytes taken from the university's student record system.
Stolen data includes passport numbers, national insurance numbers, dates of birth, financial details, ethnicity and disability information, affecting both current students and alumni. Early speculation pointed to voice phishing or a supply chain breach; reporting since has identified the entry point as an Oracle PeopleSoft zero-day, CVE-2026-35273, which Mandiant links to attacks on more than 100 organisations, most of them universities.
Security researcher Troy Hunt believes the university was held to ransom and refused to pay, which is why the data was published — the university itself won't comment while the criminal investigation runs. EMSOU's Regional Cyber Crime Unit is investigating, and the university has notified the ICO, NCSC, Office for Students and Action Fraud.
Affected individuals should enable multi-factor authentication and stay alert to unexpected phone calls.
Source: BBC News
Microsoft has attributed a supply chain attack on Mastra — an open-source TypeScript framework for building AI applications — to North Korean state actor Sapphire Sleet, also tracked as APT38 and BlueNoroff. The attribution, made June 19 with "high confidence," came after attackers compromised an npm maintainer account and poisoned over 140 packages with malicious code that ran straight from a postinstall hook.
The malware targeted cryptocurrency wallets from 166 browser extensions, including MetaMask and Coinbase Wallet, while also stealing browser history and system data. It ran on Windows, macOS and Linux, and switched off TLS certificate verification before phoning home.
Developers should check for easy-day-js dependencies. Mastra 1.13.0 and earlier and @mastra/core 1.42.0 and earlier are unaffected — anything newer needs checking.
Source: Infosecurity Magazine
Microsoft has attributed a supply chain attack on Mastra — an open-source TypeScript framework for building AI applications — to North Korean state actor Sapphire Sleet, also tracked as APT38 and BlueNoroff. The attribution, made June 19 with "high confidence," came after attackers compromised an npm maintainer account and poisoned over 140 packages with malicious code that ran straight from a postinstall hook.
The malware targeted cryptocurrency wallets from 166 browser extensions, including MetaMask and Coinbase Wallet, while also stealing browser history and system data. It ran on Windows, macOS and Linux, and switched off TLS certificate verification before phoning home.
Developers should check for easy-day-js dependencies. Mastra 1.13.0 and earlier and @mastra/core 1.42.0 and earlier are unaffected — anything newer needs checking.
Source: Infosecurity Magazine
The Education Authority (EA) wrote to parents at 23 schools in Northern Ireland on 25 June, warning that their children's personal data may have been accessed in the cyber attack first flagged to schools on 2 April. Sixteen of those schools had received no previous notification.
The attack targeted the C2K network, which handles IT for all schools across Northern Ireland, locking students out of accounts and resources during the lead-up to exam season. A 16-year-old boy was arrested and released on bail.
The EA says it still cannot confirm what personal data may have been affected, and that forensic analysis is a lengthy process. It has commissioned an independent review, notified the Information Commissioner's Office, forced a full password reset across the network, and is assisting the PSNI's criminal investigation. It reports no evidence of further breaches since the attack was detected, and says more notifications will follow as findings are confirmed.
Source: BBC News
The Education Authority (EA) wrote to parents at 23 schools in Northern Ireland on 25 June, warning that their children's personal data may have been accessed in the cyber attack first flagged to schools on 2 April. Sixteen of those schools had received no previous notification.
The attack targeted the C2K network, which handles IT for all schools across Northern Ireland, locking students out of accounts and resources during the lead-up to exam season. A 16-year-old boy was arrested and released on bail.
The EA says it still cannot confirm what personal data may have been affected, and that forensic analysis is a lengthy process. It has commissioned an independent review, notified the Information Commissioner's Office, forced a full password reset across the network, and is assisting the PSNI's criminal investigation. It reports no evidence of further breaches since the attack was detected, and says more notifications will follow as findings are confirmed.
Source: BBC News
Researchers at Unit 42 have uncovered a serious cloud attack technique called bucket hijacking, confirmed to work across Google Cloud, AWS and Microsoft Azure, all three of which were notified through responsible disclosure. The method exploits a simple but fundamental flaw: cloud storage bucket names are globally unique, meaning whoever owns the name owns the destination.
An attacker holding bucket deletion permissions inside the target environment can delete an active storage bucket, immediately re-register the same name — on Azure they'd need to be in the same tenant — and watch the original data stream of audit logs, telemetry and metrics flow silently into their environment. No alerts fire. No errors appear. The pipeline just keeps running.
Unit 42 has seen no sign of this being used in the wild, and notes that's cold comfort: once deployed, it would be extremely hard to detect. The biggest exposure is the broad storage-admin roles enterprises hand out routinely.
Unit 42 recommends restricting deletion permissions, enforcing data perimeter controls like AWS SCPs and Google Cloud VPC Service Controls, monitoring bucket deletion API calls closely, and enabling AWS account-regional S3 namespaces where available.
Source: Cybersecurity News
Researchers at Unit 42 have uncovered a serious cloud attack technique called bucket hijacking, confirmed to work across Google Cloud, AWS and Microsoft Azure, all three of which were notified through responsible disclosure. The method exploits a simple but fundamental flaw: cloud storage bucket names are globally unique, meaning whoever owns the name owns the destination.
An attacker holding bucket deletion permissions inside the target environment can delete an active storage bucket, immediately re-register the same name — on Azure they'd need to be in the same tenant — and watch the original data stream of audit logs, telemetry and metrics flow silently into their environment. No alerts fire. No errors appear. The pipeline just keeps running.
Unit 42 has seen no sign of this being used in the wild, and notes that's cold comfort: once deployed, it would be extremely hard to detect. The biggest exposure is the broad storage-admin roles enterprises hand out routinely.
Unit 42 recommends restricting deletion permissions, enforcing data perimeter controls like AWS SCPs and Google Cloud VPC Service Controls, monitoring bucket deletion API calls closely, and enabling AWS account-regional S3 namespaces where available.
Source: Cybersecurity News
A supply chain attack on market intelligence platform Klue, carried out June 11–12, has now been confirmed by roughly two dozen customers, including AlertMedia, Blackbaud, Deel, Tines, BeyondTrust and LastPass. Hackers used legacy credentials to steal OAuth tokens and bulk-exfiltrate Salesforce data. Klue has reportedly told customers the incident touches 195 of them in total.
Salesforce disabled the Klue integration on June 17 and has yet to re-enable it; Gong disabled it too.
The threat actor, Icarus, demanded ransom via a Tor leak site — but then got hacked themselves. Klue has reportedly told customers that Icarus was breached and the stolen data is now in another threat actor's hands. Klue says Icarus has begun deleting the data, and its leak site has gone dark, which suggests a ransom may have been paid. Klue hasn't confirmed either way.
Source: SecurityWeek
A supply chain attack on market intelligence platform Klue, carried out June 11–12, has now been confirmed by roughly two dozen customers, including AlertMedia, Blackbaud, Deel, Tines, BeyondTrust and LastPass. Hackers used legacy credentials to steal OAuth tokens and bulk-exfiltrate Salesforce data. Klue has reportedly told customers the incident touches 195 of them in total.
Salesforce disabled the Klue integration on June 17 and has yet to re-enable it; Gong disabled it too.
The threat actor, Icarus, demanded ransom via a Tor leak site — but then got hacked themselves. Klue has reportedly told customers that Icarus was breached and the stolen data is now in another threat actor's hands. Klue says Icarus has begun deleting the data, and its leak site has gone dark, which suggests a ransom may have been paid. Klue hasn't confirmed either way.
Source: SecurityWeek
Japan's Ground Self-Defense Force (JGSDF) unknowingly used malware-infected USB drives on classified military networks for almost a year, according to Nikkei's investigation of leaked internal documents. The counterfeit drives, manufactured in China, were distributed during earthquake relief operations in March 2024. The malware ran the instant a drive was plugged in, then sat quietly in the background.
By the time a soldier in Itami noticed his computer slowing down in February 2025, over 50 machines had connected to the infected drives — nearly half of them handling classified troop movement data. The malware matches a strain linked to a China-backed hacking group.
The JGSDF never disclosed the breach publicly, even as identical drives spread to Japanese factories and research institutions.
Source: Cybersecurity News
Japan's Ground Self-Defense Force (JGSDF) unknowingly used malware-infected USB drives on classified military networks for almost a year, according to Nikkei's investigation of leaked internal documents. The counterfeit drives, manufactured in China, were distributed during earthquake relief operations in March 2024. The malware ran the instant a drive was plugged in, then sat quietly in the background.
By the time a soldier in Itami noticed his computer slowing down in February 2025, over 50 machines had connected to the infected drives — nearly half of them handling classified troop movement data. The malware matches a strain linked to a China-backed hacking group.
The JGSDF never disclosed the breach publicly, even as identical drives spread to Japanese factories and research institutions.
Source: Cybersecurity News
Two young men convicted of the 2024 cyber-attack on Transport for London were repeat offenders well known to law enforcement long before the breach. Owen Flowers, 18, from Walsall, and Thalha Jubair, 20, from east London, pleaded guilty Monday to the attack, which disrupted TfL services for months, exposed data on millions of people and forced all 28,000 TfL staff to reset their passwords in person.
The trial heard both were part of Scattered Spider, the loose collective of young English-speaking cyber-criminals also linked to the Marks and Spencer and Co-op attacks.
Flowers first caught police attention shortly after turning 16, when officers issued a cease and desist order. Jubair has 22 previous convictions, started offending at 14, and had already received a youth rehabilitation order over Lapsus$-linked attacks on Nvidia and BT/EE. Both are wanted in the US, though the BBC understands American authorities won't pursue Flowers further; Jubair's US case involves alleged theft and extortion of $87m.
Sentencing is set for July 16. The National Crime Agency wants new Cyber Crime Risk Orders — proposed under Computer Misuse Act reforms — to let police restrict high-risk offenders before the next breach.
Source: BBC News
Two young men convicted of the 2024 cyber-attack on Transport for London were repeat offenders well known to law enforcement long before the breach. Owen Flowers, 18, from Walsall, and Thalha Jubair, 20, from east London, pleaded guilty Monday to the attack, which disrupted TfL services for months, exposed data on millions of people and forced all 28,000 TfL staff to reset their passwords in person.
The trial heard both were part of Scattered Spider, the loose collective of young English-speaking cyber-criminals also linked to the Marks and Spencer and Co-op attacks.
Flowers first caught police attention shortly after turning 16, when officers issued a cease and desist order. Jubair has 22 previous convictions, started offending at 14, and had already received a youth rehabilitation order over Lapsus$-linked attacks on Nvidia and BT/EE. Both are wanted in the US, though the BBC understands American authorities won't pursue Flowers further; Jubair's US case involves alleged theft and extortion of $87m.
Sentencing is set for July 16. The National Crime Agency wants new Cyber Crime Risk Orders — proposed under Computer Misuse Act reforms — to let police restrict high-risk offenders before the next breach.
Source: BBC News
