MCP tools are implicated in several new attack techniques. Here's a look at how they can be manipulated for good, such as logging tool usage and filtering unauthorized commands.
Over the last few months, there has been a lot of activity in the Model Context Protocol (MCP) space, both in terms of adoption as well as security. Developed by Anthropic, MCP has been rapidly gaining traction across the AI ecosystem. MCP allows Large Language Models (LLMs) to interface with tools and for those interfaces to be rapidly created. MCP tools allow for the rapid development of “agentic” systems, or AI systems that autonomously perform tasks.
Beyond adoption, new attack techniques have been shown to allow prompt injection via MCP tool descriptions and responses, MCP tool poisoning, rug pulls and more.
Prompt Injection is a weakness in LLMs that can be used to elicit unintended behavior, circumvent safeguards and produce potentially malicious responses. Prompt injection occurs when an attacker instructs the LLM to disregard other rules and do the attacker’s bidding. In this blog, I show how to use techniques similar to prompt injection to change the LLM’s interaction with MCP tools. Anyone conducting MCP research may find these techniques useful.
This report highlights a rarely-discussed but crucially important attack surface: security vendors themselves.
La France condamne avec la plus grande fermeté le recours par le service de renseignement militaire russe (GRU) au mode opératoire d’attaque APT28, à l’origine de plusieurs cyber-attaques contre des intérêts français.
Depuis 2021, ce mode opératoire d’attaque (MOA) a été utilisé dans le ciblage ou la compromission d’une dizaine d’entités françaises. Ces entités sont des acteurs de la vie des Français : services publics, entreprises privées, ainsi qu’une organisation sportive liée à l’organisation des Jeux olympiques et paralympiques 2024. Par le passé, ce mode opératoire a également été utilisé par le GRU dans le sabotage de la chaîne de télévision TV5Monde en 2015, ainsi que dans la tentative de déstabilisation du processus électoral français en 2017.
APT28 est aussi employé pour exercer une pression constante sur les infrastructures ukrainiennes dans le contexte de la guerre d’agression menée par la Russie contre l’Ukraine, notamment lorsqu’il est opéré par l’unité 20728 du GRU. De nombreux partenaires européens ont également été visés par APT28 au cours des dernières années. À ce titre, l’UE a imposé des sanctions aux personnes et entités responsables des attaques menées à l’aide de ce mode opératoire.
Europol has launched a new Operational Taskforce (OTF) to tackle the rising trend of violence-as-a-service and the recruitment of young perpetrators into serious and organised crime. Known as OTF GRIMM, the Taskforce, led by Sweden, brings together law enforcement authorities from Belgium, Denmark, Finland, France, Germany, the Netherlands, and Norway, with Europol providing operational support, threat analysis and coordination.
The exploitation of young perpetrators to carry out criminal acts has emerged as a fast-evolving tactic used by organised crime. This trend was underlined in the European Union Serious and Organised Crime Threat Assessment 2025 (EU-SOCTA), which identified the deliberate use of youngsters as a way to avoid detection and prosecution.
Violence-as-a-service refers to the outsourcing of violent acts to criminal service providers — often involving the use of young perpetrators to carry out threats, assaults, or killings for a fee.
Investigations show that these acts are often orchestrated remotely, with young people recruited and instructed online. There is a clear demand from the criminal underworld for youngsters willing to carry out violent tasks — and a supply of vulnerable young people being groomed or coerced into doing so.
Verisource Services, an employee benefits administration service provider, has determined that a previously announced data breach was far worse than initially thought and has affected up to 4 million individuals. The Houston, Texas-based company detected a hacking incident on February 28, 2024, that disrupted access to some of its systems. Third-party cybersecurity and incident response experts were engaged to investigate the incident and determine the nature and scope of the unauthorized activity.
The forensic investigation confirmed hackers had access to its network and exfiltrated files on February 27, 2024. At the time of the initial announcement, Verisource Services said names, dates of birth, genders, and Social Security numbers had been stolen. The affected individuals included employees and dependents of clients who used its services, which include HR outsourcing, benefits enrollment, billing, and administrative services.
The data breach was initially reported as affecting 1,382 individuals, but as the investigation progressed, it became clear that the breach was worse than initially thought. In August 2024, the data breach was reported to the HHS’ Office for Civil Rights (OCR) as involving the protected health information of 112,726 individuals. The most recent notification to the Maine Attorney General indicates up to 4 million individuals have been affected, a sizeable increase from previous estimates. The OCR breach portal still lists the incident as affecting 112,726 patients and plan members of its HIPAA-regulated entity clients, although that total may well be updated in the coming days.
Verisource Services explained in the breach notice that the data review was not completed until April 17, 2025, almost 14 months after the security incident was detected. Verisource Services reported the security incident to the Federal Bureau of Investigation, and several additional security measures have been implemented to improve its security posture. Notification letters had previously been sent to some affected individuals; however, the bulk of the notification letters have only recently been mailed. Verisource Services said complimentary credit monitoring and identity theft protection services have been offered to the affected individuals, who will also be protected with a $1,000,000 identity theft insurance policy.
Hitachi Vantara, a subsidiary of Japanese multinational conglomerate Hitachi, was forced to take servers offline over the weekend to contain an Akira ransomware attack.
The company provides data storage, infrastructure systems, cloud management, and ransomware recovery services to government entities and some of the world's biggest brands, including BMW, Telefónica, T-Mobile, and China Telecom.
In a statement shared with BleepingComputer, Hitachi Vantara confirmed the ransomware attack, saying it hired external cybersecurity experts to investigate the incident's impact and is now working on getting all affected systems online.
"On April 26, 2025, Hitachi Vantara experienced a ransomware incident that has resulted in a disruption to some of our systems," Hitachi Vantara told BleepingComputer.
"Upon detecting suspicious activity, we immediately launched our incident response protocols and engaged third-party subject matter experts to support our investigation and remediation process. Additionally, we proactively took our servers offline in order to contain the incident.
"We are working as quickly as possible with our third-party subject matter experts to remediate this incident, continue to support our customers, and bring our systems back online in a secure manner. We thank our customers and partners for their patience and flexibility during this time."
Following news of cyber incidents impacting UK retailers, the NCSC can confirm it is working with organisations affected.
NCSC CEO Dr Richard Horne said:
“The disruption caused by the recent incidents impacting the retail sector are naturally a cause for concern to those businesses affected, their customers and the public.
“The NCSC continues to work closely with organisations that have reported incidents to us to fully understand the nature of these attacks and to provide expert advice to the wider sector based on the threat picture.
“These incidents should act as a wake-up call to all organisations. I urge leaders to follow the advice on the NCSC website to ensure they have appropriate measures in place to help prevent attacks and respond and recover effectively.”
April 28, 2025
HALIFAX, Nova Scotia--(BUSINESS WIRE)-- Emera Inc. and Nova Scotia Power today announced, on April 25, 2025 they discovered and are actively responding to a cybersecurity incident involving unauthorized access into certain parts of its Canadian network and servers supporting portions of its business applications.
Immediately following detection of the external threat, the companies activated their incident response and business continuity protocols, engaged leading third-party cybersecurity experts, and took actions to contain and isolate the affected servers and prevent further intrusion. Law enforcement officials have been notified.
There remains no disruption to any of our Canadian physical operations including at Nova Scotia Power’s generation, transmission and distribution facilities, the Maritime Link or the Brunswick Pipeline, and the incident has not impacted the utility’s ability to safely and reliably serve customers in Nova Scotia. There has been no impact to Emera’s U.S. or Caribbean utilities.
Emera will release its Q1 Financial Statements and Management Disclosure and Analysis on May 8, 2025, as planned. At this time, the incident is not expected to have a material impact on the financial performance of the business.
Our IT team is working diligently with cyber security experts to bring the affected portions of our IT system back online.
The individuals operating under the DragonForce banner and attacking UK high street retailers are using social engineering for entry. I think it’s in the public interest to break down what is happening.
The attacks on Marks and Spencer, Co-op and Harrods are linked. DragonForce’s lovely PR team claim more are to come.
Defenders should urgently make sure they have read the CISA briefs on Scattered Spider and LAPSUS$ as it’s a repeat of the 2022–2023 activity which saw breaches at Nvidia, Samsung, Rockstar and Microsoft amongst many others. More info below.
I am not saying it is Scatter Spider; Scattered Spider has become a dumping ground for e-crime groups anyway. The point is they — the threat actor — are entering using the front door, via the helpdesk to get MFA access — those are very good guides from defenders about what to do, links below.
Source: Cybersecurity and Infrastructure Security Agency
DragonForce is a white label cartel operation housing anybody who wants to do e-crime. Some of them are pretty good at e-crime.
While organisations are away at RSA thinking about quantum AI cyber mega threats — the harsh reality is most organisations do not have the foundations in place to do be worrying about those kind of things. Generative AI is porn for execs and growth investment — threat actors are very aware that now is the time to launch attacks, not with GenAI, but foundational issues. Because nobody is paying attention.
Once they get access, they are living off the land — using Teams, Office search to find documentation, the works. Forget APTs, now you have the real threat: Advanced Persistent Teenagers, who have realised the way to evade most large cyber programmes is to cosplay as employees. Last time this happened, the MET Police ended up arresting a few under-18 UK nationals causing incidents to largely drop off.
