Full Daily Digest

A critical vulnerability in Google’s Antigravity AI agent manager has been uncovered, allowing attackers to bypass its Secure Mode sandbox and achieve remote code execution (RCE). The flaw, discovered by Pillar Security, exploited a combination of prompt injection and Antigravity’s file-creation capabilities, enabling malicious commands to execute without detection. Secure Mode, designed to isolat

CyberScoop · 2026-04-20 · Vulnerability: A critical vulnerability in Google’s Antigravity AI agent manager has been uncovered, allowing attackers to bypass its Secure Mode sandbox and achieve remote code execution (RCE).

A recent breach at Vercel highlights the growing risk of overpermissioned OAuth tokens and third-party AI tool integrations. Attackers compromised Context.ai, an AI tool vendor, and leveraged a Vercel employee's OAuth token—granted with 'Allow All' permissions—to access non-sensitive environment variables. The initial compromise stemmed from an employee downloading Roblox cheats containing an info

Dark Reading · 2026-04-20 · Incidents: A recent breach at Vercel highlights the growing risk of overpermissioned OAuth tokens and third-party AI tool integrations.

The FTC is intensifying its focus on AI-driven malicious activities, particularly nonconsensual deepfakes and voice cloning scams, leveraging the recently enacted Take It Down Act. This law enables criminal prosecution for distributing AI-generated intimate images and mandates swift content removal by platforms, with non-compliance triggering FTC investigations. The first conviction under this law

CyberScoop · 2026-04-20 · Policy: The FTC is intensifying its focus on AI-driven malicious activities, particularly nonconsensual deepfakes and voice cloning scams, leveraging the recently enacted Take It Down Act.

Serial-to-IP converters, critical for bridging legacy industrial machinery with modern networks, are riddled with vulnerabilities—both newly discovered and long-standing. Researchers uncovered 20 new flaws in these devices, alongside thousands of known vulnerabilities in their outdated software stacks, including ancient Linux kernels averaging 2,255 bugs per device. These converters, often running

Dark Reading · 2026-04-20 · Vulnerability: Serial-to-IP converters, critical for bridging legacy industrial machinery with modern networks, are riddled with vulnerabilities—both newly discovered and long-standing.

A sophisticated attack targeting Vercel originated from malware disguised as Roblox cheats, highlighting the risks of interconnected cloud systems and overprivileged SaaS integrations. The breach began in February when a Context.ai employee’s device was infected with Lumma Stealer malware, granting attackers access to AWS environments and OAuth tokens. These tokens were used to compromise a Vercel

CyberScoop · 2026-04-20 · Incidents: A sophisticated attack targeting Vercel originated from malware disguised as Roblox cheats, highlighting the risks of interconnected cloud systems and overprivileged SaaS integrations.

A critical vulnerability (CVE-2026-5760, CVSS 9.8) in SGLang, a high-performance serving framework for large language models, allows remote code execution (RCE) via malicious GGUF model files. The flaw lies in the '/v1/rerank' endpoint, where a crafted tokenizer.chat_template parameter containing a Jinja2 server-side template injection (SSTI) payload can execute arbitrary Python code. Attackers ex

The Hacker News · 2026-04-20 · Vulnerability: A critical vulnerability (CVE-2026-5760, CVSS 9.8) in SGLang, a high-performance serving framework for large language models, allows remote code execution (RCE) via malicious GGUF model files.

Italy’s data protection regulator has imposed significant fines totaling €12.5 million ($14.7 million) on Poste Italiane SpA and its subsidiary Postepay SpA for violating data privacy laws. The violations stem from excessively invasive data collection practices via their mobile apps, which monitored installed and running applications under the guise of fraud prevention. The regulator found these m

The Record by Recorded Future · 2026-04-20 · Policy: Italy’s data protection regulator has imposed significant fines totaling €12.5 million ($14.7 million) on Poste Italiane SpA and its subsidiary Postepay SpA for violating data privacy laws.

A $290 million cryptocurrency theft has been attributed to North Korean hackers, specifically the TraderTraitor group within the Lazarus operation. The attack targeted Kelp, a crypto platform, exploiting a vulnerability in LayerZero’s Decentralized Verifier Network (DVN). LayerZero, a cryptocurrency infrastructure developer, had warned against relying on a single DVN for message verification, emph

The Record by Recorded Future · 2026-04-20 · Incidents: A $290 million cryptocurrency theft has been attributed to North Korean hackers, specifically the TraderTraitor group within the Lazarus operation.

Opportunistic cyberattacks thrive on weak defaults and inconsistent security practices, a challenge Microsoft addresses through its Secure Future Initiative (SFI) and One Microsoft principles. Ilya Grebnov, Deputy CISO for Microsoft Dynamics 365 and Power Platform, emphasizes the importance of standardization, paved paths, and platform-managed identities to reduce attack surfaces. By enforcing con

Microsoft Security Blog · 2026-04-20 · Vulnerability: Opportunistic cyberattacks thrive on weak defaults and inconsistent security practices, a challenge Microsoft addresses through its Secure Future Initiative (SFI) and One Microsoft principles.

Elon Musk’s failure to appear for questioning by French police highlights the escalating legal and regulatory scrutiny facing X (formerly Twitter) over allegations of distributing child sexual abuse material (CSAM) and generating non-consensual sexualized AI images. The investigation, initiated after raids on X’s Paris offices in February, aims to enforce compliance with French law rather than imp

The Record by Recorded Future · 2026-04-20 · Incidents: Elon Musk’s failure to appear for questioning by French police highlights the escalating legal and regulatory scrutiny facing X (formerly Twitter) over allegations of distributing child sexual abuse…

WhatsApp's metadata leakage exposes user activity and device information to potential attackers without requiring direct interaction or sophisticated exploits. Tal Be'ery demonstrated how a simple custom program can silently ping a target's device, revealing online status and device type using only a phone number. This flaw, rooted in WhatsApp's design choices, enables both nation-state actors and

Dark Reading · 2026-04-20 · Vulnerability: WhatsApp's metadata leakage exposes user activity and device information to potential attackers without requiring direct interaction or sophisticated exploits.

A cyberattack on France's National Agency for Secure Documents (ANTS) has potentially exposed sensitive personal data, including login credentials, names, email addresses, dates of birth, and unique identifiers. The breach, detected on April 15, may also have compromised postal addresses, places of birth, and phone numbers. While uploaded documents remain secure, the incident underscores vulnerabi

The Record by Recorded Future · 2026-04-20 · Incidents: A cyberattack on France's National Agency for Secure Documents (ANTS) has potentially exposed sensitive personal data, including login credentials, names, email addresses, dates of birth, and unique…

Bluesky experienced a significant outage due to a sophisticated DDoS attack, disrupting core functionalities like feeds, notifications, and search. The attack, which began on April 15, was mitigated by engineers overnight, though the platform remained under ongoing assault. While Bluesky confirmed no unauthorized access to user data, the Iran-linked 313 Team claimed responsibility, citing retaliat

The Record by Recorded Future · 2026-04-20 · Incidents: Bluesky experienced a significant outage due to a sophisticated DDoS attack, disrupting core functionalities like feeds, notifications, and search.

Recent incidents highlight a troubling trend where attackers exploit trusted third-party tools and workflows to gain unauthorized access. The Vercel breach, originating from a compromised AI tool (Context.ai), underscores how supply chain vulnerabilities can escalate into full-blown system compromises. Attackers are increasingly leveraging multi-stage payloads, memory-resident code, and legitimate

The Hacker News · 2026-04-20 · Incidents: Recent incidents highlight a troubling trend where attackers exploit trusted third-party tools and workflows to gain unauthorized access.

A suspected North Korean threat actor recently compromised the widely-used Axios JavaScript library, injecting malicious code that could have impacted millions of downloads across enterprises, startups, and government systems. The attack was detected within minutes by an AI-powered monitoring tool, highlighting the critical role of AI in defending against supply chain threats. The rapid response—r

CyberScoop · 2026-04-20 · Incidents: A suspected North Korean threat actor recently compromised the widely-used Axios JavaScript library, injecting malicious code that could have impacted millions of downloads across enterprises, startups, and…

Cloudflare’s approach to scaling AI-driven code review highlights both the potential and limitations of integrating AI into software development workflows. By leveraging a multi-agent system that specializes in areas like security, performance, and compliance, Cloudflare has significantly reduced review bottlenecks while maintaining cost efficiency. However, the system faces challenges such as lac

Cloudflare Blog · 2026-04-20 · Tools: Cloudflare’s approach to scaling AI-driven code review highlights both the potential and limitations of integrating AI into software development workflows.

Cloudflare has successfully integrated AI into its engineering stack, leveraging internally built infrastructure to enhance developer productivity and code quality. Over the past eleven months, the company formed a dedicated team, iMARS, to develop MCP servers, AI tooling, and an access layer tailored for internal use. The results are striking: 93% of Cloudflare’s R&D organization now actively use

Cloudflare Blog · 2026-04-20 · Tools: Cloudflare has successfully integrated AI into its engineering stack, leveraging internally built infrastructure to enhance developer productivity and code quality.

Cloudflare’s Agents Week 2026 highlights the transformative role of AI agents in reshaping cloud infrastructure and web interactions. As agents become integral to workflows—ranging from coding and support to research—the demand for scalable, secure, and agent-optimized compute platforms has surged. Cloudflare’s serverless Workers platform, initially launched eight years ago, has evolved to meet th

Cloudflare Blog · 2026-04-20 · Events: Cloudflare’s Agents Week 2026 highlights the transformative role of AI agents in reshaping cloud infrastructure and web interactions.

A British national, Tyler Robert Buchanan, has pleaded guilty in the US for his role in the financially motivated hacking group Scattered Spider, admitting to SMS phishing attacks and stealing millions in cryptocurrency. Buchanan and his co-conspirators used phishing kits to harvest credentials and PII, then accessed victim systems to steal sensitive data, including intellectual property and confi

SecurityWeek · 2026-04-20 · Incidents: A British national, Tyler Robert Buchanan, has pleaded guilty in the US for his role in the financially motivated hacking group Scattered Spider, admitting to SMS phishing attacks and stealing millions in…

Threat actors are increasingly leveraging QEMU, a cross-platform machine emulator, to evade detection and deploy ransomware and remote access tools. Sophos reports a surge in malicious campaigns since late 2025, with attackers using QEMU to establish covert reverse SSH backdoors, harvest credentials, and maintain persistence. In one campaign, STAC4713, linked to the PayoutsKing ransomware, hackers

SecurityWeek · 2026-04-20 · Incidents: Threat actors are increasingly leveraging QEMU, a cross-platform machine emulator, to evade detection and deploy ransomware and remote access tools.

AI deployments often falter after the initial demo phase due to a stark mismatch between controlled demo environments and real-world operational complexities. While demos showcase potential with clean data and ideal scenarios, production environments introduce messy data, inconsistent inputs, and fragmented systems, leading to performance degradation. Security and IT teams face specific challenges

The Hacker News · 2026-04-20 · Vulnerability: AI deployments often falter after the initial demo phase due to a stark mismatch between controlled demo environments and real-world operational complexities.

Bluesky, a decentralized microblogging platform, faced a sophisticated DDoS attack starting late on April 15, causing intermittent service disruptions for nearly 24 hours. The attack targeted critical functionalities like feeds, notifications, and search, though no unauthorized access to user data was detected. While the pro-Iran hacktivist group 313 Team claimed responsibility, their assertions r

SecurityWeek · 2026-04-20 · Incidents: Bluesky, a decentralized microblogging platform, faced a sophisticated DDoS attack starting late on April 15, causing intermittent service disruptions for nearly 24 hours.

A critical design flaw in Anthropic's Model Context Protocol (MCP) architecture has been uncovered, enabling remote code execution (RCE) across systems utilizing its SDK. This vulnerability, inherent in the protocol's STDIO interface, allows attackers to execute arbitrary commands, compromising sensitive data, internal databases, and API keys. Over 7,000 publicly accessible servers and 150 million

The Hacker News · 2026-04-20 · Vulnerability: A critical design flaw in Anthropic's Model Context Protocol (MCP) architecture has been uncovered, enabling remote code execution (RCE) across systems utilizing its SDK.

The U.S. Senate has extended Section 702 of the Foreign Intelligence Surveillance Act (FISA) until April 30, following a chaotic legislative process that highlighted deep divisions over privacy and national security. This controversial program allows intelligence agencies to collect overseas communications without warrants, often incidentally capturing Americans' data. While officials argue it's c

SecurityWeek · 2026-04-20 · Policy: The U.S. Senate has extended Section 702 of the Foreign Intelligence Surveillance Act (FISA) until April 30, following a chaotic legislative process that highlighted deep divisions over privacy and national…

Frontier AI models are reshaping the cybersecurity landscape by autonomously identifying and exploiting software vulnerabilities at an unprecedented scale and speed. These models, capable of reasoning like full-spectrum security researchers, pose significant risks by collapsing the patching window for N-day vulnerabilities and enabling real-time adaptation to bypass hardened controls. Open-source

Palo Alto Unit 42 · 2026-04-20 · Vulnerability: Frontier AI models are reshaping the cybersecurity landscape by autonomously identifying and exploiting software vulnerabilities at an unprecedented scale and speed.

Network background noise, particularly spikes in traffic targeting specific vendors, can serve as an early-warning system for impending edge-device vulnerabilities, according to GreyNoise research. During a 103-day study, roughly half of the detected activity surges were followed by a vulnerability disclosure from the targeted vendor within three weeks, with a median warning period of nine days. T

CyberScoop · 2026-04-20 · Vulnerability: Network background noise, particularly spikes in traffic targeting specific vendors, can serve as an early-warning system for impending edge-device vulnerabilities, according to GreyNoise research.

A recent Censys report highlights that nearly half of the 6 million internet-facing FTP servers lack encryption, exposing organizations and users to significant security risks. FTP, a protocol in use for over 50 years, transmits data in cleartext, making it inherently insecure. While the number of FTP hosts has decreased by 40% since 2024, 2.45 million servers still show no evidence of encryption,

SecurityWeek · 2026-04-20 · Vulnerability: A recent Censys report highlights that nearly half of the 6 million internet-facing FTP servers lack encryption, exposing organizations and users to significant security risks.

The Vercel April 2026 breach underscores the critical importance of scrutinizing all environment variables, not just those marked as sensitive. The incident began with the compromise of Context.ai, a third-party AI tool, where an attacker hijacked a Vercel employee's Google Workspace OAuth app. This allowed access to non-sensitive environment variables, prompting Vercel to advise customers to rota

GitGuardian Blog · 2026-04-20 · Incidents: The Vercel April 2026 breach underscores the critical importance of scrutinizing all environment variables, not just those marked as sensitive.

A sophisticated phishing campaign targeting iOS users has resurfaced, leveraging fake crypto wallet apps in the App Store to steal recovery phrases and private keys. Discovered by Kaspersky in March 2026, these malicious apps mimic popular wallets like Metamask and Coinbase, redirecting users to fraudulent sites distributing trojanized versions. The campaign exploits regional restrictions in China

Kaspersky Securelist · 2026-04-20 · Incidents: A sophisticated phishing campaign targeting iOS users has resurfaced, leveraging fake crypto wallet apps in the App Store to steal recovery phrases and private keys.

A recent SEC enforcement action against a national securities firm underscores the growing importance of cybersecurity as a strategic business risk rather than a purely technical concern. The firm’s failure to implement basic security measures like multi-factor authentication and incident response plans led to a breach exposing 8,500 individuals’ personal information, resulting in a $325,000 penal

Cybersecurity Dive · 2026-04-20 · Case Studies: A recent SEC enforcement action against a national securities firm underscores the growing importance of cybersecurity as a strategic business risk rather than a purely technical concern.

Vercel, the creator of Next.js and a prominent frontend cloud platform, recently confirmed a significant security breach after a hacker attempted to sell stolen data, including databases, access keys, employee accounts, and source code, for $2 million. The breach originated from a compromise of Context.ai, a third-party AI tool used by a Vercel employee, which allowed the attacker to gain access t

SecurityWeek · 2026-04-20 · Incidents: Vercel, the creator of Next.js and a prominent frontend cloud platform, recently confirmed a significant security breach after a hacker attempted to sell stolen data, including databases, access keys,…

A critical command injection vulnerability (CVE-2023-33538, CVSS 8.8) in discontinued TP-Link routers has been under active but unsuccessful exploitation attempts for over a year. The flaw, rooted in improper sanitization of the ssid1 parameter in HTTP GET requests, affects several TL-WR series models (v1/v2/v4/v8/v10), allowing authenticated attackers to execute arbitrary commands. Despite public

SecurityWeek · 2026-04-20 · Vulnerability: A critical command injection vulnerability (CVE-2023-33538, CVSS 8.8) in discontinued TP-Link routers has been under active but unsuccessful exploitation attempts for over a year.

A newly discovered malware, ZionSiphon, has been identified targeting Israeli water treatment and desalination systems, raising alarms about the increasing sophistication of politically motivated attacks on critical infrastructure. The malware, detected in late June 2025, leverages privilege escalation, persistence mechanisms, and USB propagation to infiltrate operational technology (OT) systems.

The Hacker News · 2026-04-20 · Incidents: A newly discovered malware, ZionSiphon, has been identified targeting Israeli water treatment and desalination systems, raising alarms about the increasing sophistication of politically motivated attacks on…

A sophisticated breach at Vercel, linked to the compromise of Context.ai, highlights the risks of third-party tool integrations and overly permissive OAuth configurations. Attackers leveraged a Context.ai employee's Lumma Stealer infection to pivot into Vercel's infrastructure, accessing non-sensitive environment variables and a subset of customer credentials. While encrypted 'sensitive' variables

The Hacker News · 2026-04-20 · Incidents: A sophisticated breach at Vercel, linked to the compromise of Context.ai, highlights the risks of third-party tool integrations and overly permissive OAuth configurations.