Agenti OpenAI, è escalation di incidenti: che succede ora
OpenAI ha avvisato oltre 100 organizzazioni potenzialmente impattate in incidenti con suoi agenti AI fuori controllo, mentre la California ha appena aperto un’indagine. Da questione solo tecnica e politica, il caso monta e si appresta a diventare giudiziario. Con conseguenze che si dipaneranno nei prossimi mesi
US sanctions Tren de Aragua gang members in ATM hacks crackdown
The U.S. Treasury Department has sanctioned eight members of the Venezuelan gang Tren de Aragua (TdA) for their role in the theft of millions of dollars in ATM jackpotting attacks across the United States.
Se l'approccio nella gestione di un incidente è la ricerca del colpevole, probabilmente non ci sarà alcuna lezione appresa e anzi l'organizzazione esprimerà, prima o poi, la propria colpa con sintomi quali il reiterarsi dell'evento o l'aggravarsi delle conseguenze. Con tutte le conseguenze finanziarie, reputazionali, legali, operative e strategiche
Browser-based attacks can steal sessions, abuse extensions, or manipulate users without creating the endpoint artifacts EDR is designed to detect. NordLayer explains three ways attacks can evade endpoint telemetry and why browser-level controls can help close the gap.
Custom GPT malevoli: così ChatGPT diventa il primo anello di un attacco ClickFix
Un Custom GPT ospitato su ChatGPT convince le vittime a raggiungere una falsa verifica di sicurezza che avvia una catena ClickFix fino all’installazione di un trojan ad accesso remoto (RAT). La campagna mostra come gli attaccanti stiano trasformando la fiducia nelle piattaforme AI e nei servizi legittimi in una nuova superficie d’attacco
Dell asks admins to patch max severity CSM flaws as soon as possible
Dell has patched two maximum severity vulnerabilities in the Container Storage Modules (CSM) that connect Dell enterprise storage arrays to Kubernetes environments.
Microsoft’s X account hacked in crypto pump-and-dump scheme
On Thursday, unknown attackers hijacked the official Microsoft account on X, which has over 13 million followers, in what appeared to be a pump-and-dump scheme promoting a crypto token.
Trattamenti per finalità di Polizia: come governare la circolazione dei dati
Quando un dato passa dalla Polizia Locale alle Forze di Polizia non cambia soltanto destinatario: può entrare in un diverso sistema di responsabilità. Distinguere movimento interno e comunicazione è il primo passo per governare base giuridica, canali, contenuto, tracciabilità e gestione degli errori
Anatomía de BraZetsu: Cómo los cibercriminales abastecen el ecosistema clandestino
Group-IB descubre BraZetsu, un nuevo malware para Windows basado en Python que funciona como un toolkit maestro para Initial Access Brokers y potencia un marketplace clandestino único, mejorado con IA, para comercializar objetivos comprometidos en Iberoamérica y Latinoamérica.
Fortinet warns of critical FortiMail flaw exploited in zero-day attacks
Fortinet is warning customers of a critical FortiMail vulnerability, tracked as CVE-2026-104286, that is being actively exploited in zero-day attacks to execute unauthorized code or commands on vulnerable devices.
Microsoft says threat actors are ahead in the early AI race
Microsoft says cyberattackers are currently benefiting from artificial intelligence faster than defenders, allowing threat actors to speed up vulnerability discovery, malware development, and post-compromise activity while security teams struggle to keep pace.
In this week’s edition, Amy reflects on the importance of prioritizing family and personal well-being over the pressure to remain constantly productive.
Microsoft Digital Defense Report 2026: l’AI accelera gli attacchi e amplia il perimetro
Il Microsoft Digital Defense Report 2026 fotografa un rischio cyber sempre più interconnesso: l’AI accelera gli attacchi, l’identità resta il principale punto di controllo e gli agenti aprono una nuova superficie da proteggere. L’Italia è quinta in Europa per attività osservata
Agenti di OpenAI: nascoste le attività di hacking nelle violazioni governative USA
Modelli di OpenAI hanno estratto dati da 55 siti web, fra cui la SEC, la Consob americana. Alla persistenza dell’obiettivo, all'esplorazione autonoma dell’ambiente e all'aggiramento degli ostacoli, ora si affianca la riduzione della tracciabilità delle azioni. Ecco cosa emerge dall'analisi delle tattiche inedite degli agenti di OpenAI nell'hacking
Meta AI Glasses: tanti i rischi privacy, ecco perché
L’autorità privacy di Amburgo ha sottoposto i Ray-Ban Meta AI Glasses a un’analisi tecnica e giuridica che solleva dubbi sulla liceità delle riprese di terzi, sulla trasparenza e sull’uso dei dati per il training dell’AI. E per chi indossa gli occhiali gli obblighi GDPR possono essere molto più ampi di quanto sembri
Police dismantle KillSec ransomware gang allegedly led by 16-year-old
An international law enforcement operation dubbed "Operation KillSwitch" seized the KillSec ransomware gang's data leak site and servers, led to three arrests, and identified a 16-year-old as the group's alleged administrator.
Zero Trust can verify users once they are established, but onboarding creates a gap where organizations must decide who to trust before strong authentication exists. Specops explains why identity verification should begin before credentials, MFA methods, and access are issued.
No files with concatenated file extensions pointing to a particular ransomware. In both cases the victim organization — one in healthcare, one in real estate — learned it had a problem in similar ways: an extortion message demanding payment within 72 hours, sent from inside its own Microsoft 365 tenant. The uncomfortable takeaway: in the cloud, an attacker doesn't need ransomware to run a ransomware operation. Why This Matters: Ransomware Is a Category, not a Payload It's tempting to file this under "not really ransomware" because nothing got encrypted. Ransomware has never actually been about encryption — encryption is just one tool for the thing that pays: extortion. And there is very little to "bypass," because gathering files through the Microsoft Graph API isn't an exploit — it's the platform doing exactly what it was built to do for an authenticated user. In the cloud, the question stops being "what did the malware do" and becomes "who is holding this token, and should they be." Key Takeaways Ransomware is the extortion, not the encryption. The actor gets ransomware outcomes — a ransom demand, a ticking clock, real leverage — with no malware, no encryption, and no file-extension changes. The Microsoft Graph API lets an authenticated attacker enumerate and pull mail, SharePoint, and OneDrive files at scale — nothing to install and no security controls to defeat, because its sanctioned functionality being used as designed. Credential & Session Theft: Adversary-in-the-Middle The look-alike portal isn't just a credential grabber — it's a reverse proxy sitting between the victim and the real identity provider (Microsoft Entra ID, Okta, etc.). With a live session token, the attacker is inside as the user, and MFA has already been satisfied. In the Real Estate case, the audit log tells the story in about ninety seconds — an Update user operation, a POST to UserAuthMethod.SoftwareOathProofupRegistration, and a User registered security info event, all tying a new OATH/TOTP software token to the victim's account. According to a recent article by Microsoft, the Update user operation contained a record of the new software token being added with the “…device name NO_DEVICE, the device token NO_DEVICE_TOKEN, and the SoftwareTokenActivated device tag.” We urge organizations threat hunt for similar TTPs in their environments. Programmatic Exfiltration: The Graph API Is the Easy Button With durable access established, the operators stop clicking and start scripting. Across both cases we saw a dedicated source IP performing non-interactive sign-ins — token-based authentication with no interactive prompt — driving automated collection through the Microsoft Graph API against Exchange Online, SharePoint, and OneDrive. The attacker isn't exploiting a vulnerability or defeating a control — it's calling the same API Microsoft publishes, as an authenticated user the platform already trusts. One investigation parsed tens of thousands of mailbox items, and, in SharePoint alone, over 350,000 file access events. If your alerting relies on the user noticing a "new device added" email, that email may be gone before they see it. Expanding the Blast Radius: Guests, SSO, and Business Email A compromised Microsoft 365 identity is rarely just one mailbox. The Payload That Isn't: Extortion Without Encryption This is where the ransomware comparison stops being a metaphor. In the Real Estate case, the actor used the compromised account to blast an email to 25 internal recipients and to post a Microsoft Teams message — both titled "YOU HAVE 72 HOURS TO CONTACT US." That Teams message was the first thing anyone noticed. Two Cases, One Operator The two AOD engagements share a single fingerprint: an attacker registered MFA method for persistence, a dedicated non-interactive source IP driving Graph API exfiltration, deletion of notification emails to blind the victim, and a 72-hour extortion demand delivered from inside the tenant. The AiTM proxy captures a session that has already cleared MFA, so the second factor is satisfied by the real user in real time.
