Sicurezza IT e IA

Blog

Ricerca, sicurezza informatica, diritto e privacy nell’assistenza: fonti originali, limiti e raccomandazioni pratiche.

Articoli consigliati

Ricerca IAAltaIn evidenza

Sleeper Agents: What a study reveals about hidden backdoors in AI

Deliberately trained backdoors can survive subsequent safety training. Understanding the experimental setup is essential to interpreting the result.

Leggi di più →
Ricerca IAAltaIn evidenza

Nightshade: How altered images can influence image-model training

Nightshade investigates targeted training-data poisoning. Its effect depends on which images enter training and how that training is configured.

Leggi di più →
Ricerca IAAltaIn evidenza

Adversarial suffixes: When text appendices change AI safety behaviour

Research from 2023 demonstrated attacks transferable between several tested models. It did not establish a master key for every AI system.

Leggi di più →
Ricerca IAAltaIn evidenza

Indirect Prompt Injection: When outside content becomes an unwanted instruction

Documents and messages can contain instructions that influence an assistant. Tools, permissions and approval steps determine the potential impact.

Leggi di più →
Ricerca IAAltaIn evidenza

Morris II: A research demonstration of self-propagating AI instructions

The research describes an attack chain between connected AI applications. It depends on specific retrieval and forwarding behaviour.

Leggi di più →
Minacce IAAltaIn evidenza

Deepfake fraud in Hong Kong: What the HK$200 million case establishes

An official account describes a prerecorded conference followed by messaging instructions, rather than a fully interactive meeting of digital doubles.

Leggi di più →
Minacce IAAltaIn evidenza

Voice cloning: How a familiar voice can support an emergency scam

Recognising a voice does not establish identity. Verify the claimed emergency through an independent channel.

Leggi di più →
Minacce IAAltaIn evidenza

Loading AI models: Why unknown pickle files can execute code

Python explicitly warns against untrusted pickle data. Review the file format and any accompanying model software separately.

Leggi di più →
Diritto e leggiAltaIn evidenza

Prohibited AI practices: Why the actual use matters

The AI Act prohibits specified practices with defined conditions. Labels such as social scoring are not a substitute for examining the actual system.

Leggi di più →
Diritto e leggiAltaIn evidenza

AI Act transparency: Distinguishing obligations applying from August 2026

Article 50, general-purpose model rules, technical detectability and visible disclosure concern different requirements.

Leggi di più →
Diritto e leggiAltaIn evidenza

NIS-2 in Germany: Scope and staged incident reporting

The BSIG separates covered entities, risk management and reporting stages. A headline company count does not determine your own status.

Leggi di più →
Sentenze chiaveAltaIn evidenza

BGH on loss of control: When a data breach can cause non-material damage

VI ZR 10/24 recognises possible damage from loss of control, not automatic compensation or a fixed tariff.

Leggi di più →
Case di curaAltaIn evidenza

Cameras in care rooms: Consider images, audio and everyone affected

Well-intended monitoring can intrude deeply into privacy. Criminal liability and wider lawfulness are separate questions.

Leggi di più →
Case di curaAltaIn evidenza

Ransomware in care: Prepare availability and recovery

A backup is not a tested continuity process. Essential information needs an appropriate route back into use.

Leggi di più →
Assistenza ambulatorialeAltaIn evidenza

Messaging in care: Secure communication needs more than encrypted content

Patient information, contacts, devices and recipients need an agreed workflow.

Leggi di più →
Assistenza ambulatorialeAltaIn evidenza

Care documentation: Traceable correction is different from falsification

Original entries and change timing must remain recognisable. Employment consequences require a separate case-specific assessment.

Leggi di più →
Sanitaria / ortopediaAltaIn evidenza

Medical supply businesses and TI: Plan connection and digital prescriptions separately

Connecting to the infrastructure does not mean every prescription workflow is already operational.

Leggi di più →
Consulenza assistenzialeAltaIn evidenza

Video care consultations under section 37 SGB XI: Check requirements, not brand slogans

Eligibility, first home visit, time limit and technical requirements must be considered together.

Leggi di più →
Assistenza ambulatorialeAltaIn evidenza

TI Messenger and WhatsApp Business: Compare more than encryption

TI Messenger provides a governed healthcare messaging framework. Approval does not replace endpoint management, recipient checks or documentation.

Leggi di più →
Ricerca IAAltaIn evidenza

AI dictation at the bedside: Recording authority, data flow and clinical review

Dictating notes differs from recording a patient conversation. Assess recording authority, processing and human approval before introducing AI notes.

Leggi di più →
Diritto e leggiAltaIn evidenza

Healthcare cloud and section 393 SGB V: Assess scope and C5 evidence precisely

Section 393 sets specific cloud conditions and exceptions. A German server location or an arbitrary ISO certificate is not a complete assessment.

Leggi di più →
Assistenza ambulatorialeAltaIn evidenza

Care-sector cloud backups: Security depends on tested restoration

Synchronisation, version history and backups serve different purposes. Plan independent copies and verify that the care application can actually restart.

Leggi di più →
Assistenza ambulatorialeAltaIn evidenza

The ePA in care: Authorised access, not a complete record at one click

The ePA can support care information exchange. Understand access rights, objections, missing documents and the facility’s own documentation responsibilities.

Leggi di più →

Tutti i temi

Ricerca IAAltaIn evidenza

Sleeper Agents: What a study reveals about hidden backdoors in AI

Deliberately trained backdoors can survive subsequent safety training. Understanding the experimental setup is essential to interpreting the result.

Leggi di più →
Ricerca IAAltaIn evidenza

Nightshade: How altered images can influence image-model training

Nightshade investigates targeted training-data poisoning. Its effect depends on which images enter training and how that training is configured.

Leggi di più →
Ricerca IA

Glaze: Making AI imitation of an artist’s style harder

Glaze modifies artwork to interfere with AI style mimicry. Its tested scope matters more than promises of invisible, permanent protection.

Leggi di più →
Ricerca IAAltaIn evidenza

Adversarial suffixes: When text appendices change AI safety behaviour

Research from 2023 demonstrated attacks transferable between several tested models. It did not establish a master key for every AI system.

Leggi di più →
Ricerca IAAltaIn evidenza

Indirect Prompt Injection: When outside content becomes an unwanted instruction

Documents and messages can contain instructions that influence an assistant. Tools, permissions and approval steps determine the potential impact.

Leggi di più →
Ricerca IAAltaIn evidenza

Morris II: A research demonstration of self-propagating AI instructions

The research describes an attack chain between connected AI applications. It depends on specific retrieval and forwarding behaviour.

Leggi di più →
Ricerca IA

Training-data extraction: When language models repeat learned text

A GPT-2 study demonstrated verbatim recovery of individual training examples. That is different from freely searching an entire training dataset.

Leggi di più →
Ricerca IA

Repetition and data extraction: Understanding a historical chatbot attack

A 2023 study investigated extraction from production models. The well-known repetition experiment needs technical and historical context.

Leggi di più →
Ricerca IA

Web poisoning: Why training-data provenance matters

Mutable web sources can threaten dataset integrity. Manipulating a dataset and demonstrating a particular model failure are separate steps.

Leggi di più →
Ricerca IA

Jailbroken: Explanations for failures of AI safety training

Competing objectives and generalisation gaps help explain tested failures. They are not a complete mathematical account of all jailbreaks.

Leggi di più →
Ricerca IAAlta

Many-Shot Jailbreaking: How numerous example dialogues influence safety behaviour

Anthropic’s 2024 study used fabricated dialogues with disallowed answers, not a sequence of harmless stories.

Leggi di più →
Ricerca IA

Sycophancy: When AI agrees too readily with a user

Confident agreement is not evidence of truth. Research investigates how preference feedback can shape an assistant’s responses.

Leggi di più →
Ricerca IA

Fine-tuning: Why safety behaviour needs to be reassessed

Research found both deliberate and unintended safety degradation after additional training. Ten examples are one experimental result, not a universal threshold.

Leggi di più →
Ricerca IAAlta

AI-generated spearphishing: What a participant study shows

Personalised deceptive messages can be automated. Experimental click rates must be distinguished from account compromise and universal success.

Leggi di più →
Ricerca IA

Model inversion: What prediction responses can reveal

Research reconstructed recognisable facial images using confidence values. Preconditions and reconstruction quality are crucial to interpreting the result.

Leggi di più →
Minacce IAAltaIn evidenza

Deepfake fraud in Hong Kong: What the HK$200 million case establishes

An official account describes a prerecorded conference followed by messaging instructions, rather than a fully interactive meeting of digital doubles.

Leggi di più →
Minacce IAAltaIn evidenza

Voice cloning: How a familiar voice can support an emergency scam

Recognising a voice does not establish identity. Verify the claimed emergency through an independent channel.

Leggi di più →
Minacce IAAltaIn evidenza

Loading AI models: Why unknown pickle files can execute code

Python explicitly warns against untrusted pickle data. Review the file format and any accompanying model software separately.

Leggi di più →
Minacce IA

Shadow AI: Confidential information in unapproved services

Unapproved use obscures data flows. Processing an input, retaining it and training on it are separate questions.

Leggi di più →
Minacce IAAlta

Slopsquatting: When AI recommends software packages that do not exist

Invented package names can create a supply-chain opportunity. A hallucination and confirmed malware are not the same finding.

Leggi di più →
Minacce IAAlta

Quishing: Checking phishing links concealed in QR codes

A QR code is a way to carry an address, not proof that the destination is trustworthy.

Leggi di più →
Minacce IAAlta

AI tools: Restrict permissions instead of relying on trust

Changing files or sending messages requires an enabling integration. Unnecessary capabilities magnify mistakes and manipulation.

Leggi di più →
Minacce IAAlta

Video identification: A visible face is only part of verification

Assess the complete regulated procedure rather than judging identity from a convincing video alone.

Leggi di più →
Minacce IA

Image-based disclosure: Why the chat interface needs security review

An image address can carry information, but disclosure requires both embedding the information and making the external request.

Leggi di più →
Minacce IA

RAG: How outside documents can influence a company assistant

Retrieval does not automatically make content trustworthy. Incorrect information and concealed instructions are different risks.

Leggi di più →
Minacce IA

AI and CAPTCHA: What research on reCAPTCHA v2 actually tested

Results for a specialised image-solving approach do not describe every form of bot protection.

Leggi di più →
Minacce IA

BlackMamba: A manufacturer’s experiment with AI-generated malicious functionality

HYAS reported a 2023 proof of concept using runtime-generated code. It does not establish failure of every endpoint product.

Leggi di più →
Minacce IAAlta

CEO fraud with impersonated media: Check identity and authority separately

Messaging and altered media can support payment fraud. A convincing appearance does not replace a traceable approval.

Leggi di più →
Minacce IAAlta

AI and erasure: Separate training data, models and document stores

Removing personal information is not one technical operation. The relevant data layer and processing purpose matter.

Leggi di più →
Minacce IA

AI Cyber Challenge: Automated vulnerability discovery with documented results

DARPA’s corrected 2025 results distinguish synthetic flaws, real findings and proposed patches.

Leggi di più →
Diritto e leggiAltaIn evidenza

Prohibited AI practices: Why the actual use matters

The AI Act prohibits specified practices with defined conditions. Labels such as social scoring are not a substitute for examining the actual system.

Leggi di più →
Diritto e leggiAltaIn evidenza

AI Act transparency: Distinguishing obligations applying from August 2026

Article 50, general-purpose model rules, technical detectability and visible disclosure concern different requirements.

Leggi di più →
Diritto e leggiAltaIn evidenza

NIS-2 in Germany: Scope and staged incident reporting

The BSIG separates covered entities, risk management and reporting stages. A headline company count does not determine your own status.

Leggi di più →
Diritto e leggiAlta

NIS-2 management duties: Implementation, oversight and training

Section 38 BSIG specifies management responsibilities. Liability still depends on the applicable legal conditions.

Leggi di più →
Diritto e leggiAlta

Cyber Resilience Act: Product security and phased obligations

The CRA concerns products with digital elements. CE marking is not a promise that a product can never be compromised.

Leggi di più →
Diritto e leggi

DORA: Organising digital resilience in finance

DORA connects risks, incidents, tests and third-party services. Advanced penetration-testing duties do not apply identically to every business.

Leggi di più →
Diritto e leggi

EU-US Data Privacy Framework: Decisions, judgments and recipient checks

Transfer assessment needs the actual recipient and legal basis, not speculation about a future court outcome.

Leggi di più →
Diritto e leggiAlta

Automated decisions: Understanding Article 22 GDPR

Not every automated calculation is covered. Sole automation and legal or similarly significant effects are central conditions.

Leggi di più →
Diritto e leggi

Garante and ChatGPT in 2023: Understanding the historical measure

The authority’s intervention illustrates legal basis and transparency issues. It must be separated from later proceedings and current settings.

Leggi di più →
Diritto e leggi

Data protection officers: Appointment, independent advice and responsibility

Specific provisions define the role. An unidentified legislative reform is not an adequate source.

Leggi di più →
Diritto e leggi

Data Act: Access to connected-product data is not universal data ownership

The Data Act strengthens certain access rights while retaining boundaries concerning personal data and trade secrets.

Leggi di più →
Diritto e leggi

AI liability: The withdrawn directive proposal is not applicable law

The proposal was withdrawn in October 2025. Its intended evidential rules do not create rights through that unadopted directive.

Leggi di più →
Diritto e leggiAlta

Meta AI training: What the DPC reported in May 2025

The 2024 pause was not the last documented development. The Irish authority later described revised plans and safeguards.

Leggi di più →
Diritto e leggi

AI at work: Assessing works-council involvement

Actual functions and deployment matter more than the name of an office extension.

Leggi di più →
Diritto e leggi

Cookie banners: Meaningful choices instead of manipulative design

Necessary functions and consent-dependent access differ. A banner needs to reflect actual technical behaviour.

Leggi di più →
Sentenze chiaveAltaIn evidenza

BGH on loss of control: When a data breach can cause non-material damage

VI ZR 10/24 recognises possible damage from loss of control, not automatic compensation or a fixed tariff.

Leggi di più →
Sentenze chiaveAlta

CJEU C-340/21: Assess attacks, safeguards and damage separately

A successful attack proves neither inadequate safeguards automatically nor automatic exemption from responsibility.

Leggi di più →
Sentenze chiaveAlta

CJEU C-300/21: No seriousness threshold, but damage is still required

The Österreichische Post judgment distinguishes infringement, damage and causation. Minor damage is different from no damage.

Leggi di più →
Sentenze chiaveAlta

SCHUFA scoring: When a score can itself be an automated decision

C-634/21 concerns the score’s determining role in subsequent decisions, not an unconditional ban on credit calculations.

Leggi di più →
Sentenze chiave

CJEU C-184/20: Public transparency needs personal-data boundaries

The Lithuanian declarations case addresses necessity and sensitive inferences, not a general ban on public transparency.

Leggi di più →
Sentenze chiaveAlta

BGH XI ZR 107/22: Proving payment authorisation

The March 2024 judgment concerns burden of proof under historical payment law, not a universal ruling on phishing negligence.

Leggi di più →
Diritto e leggi

Push-TAN fraud: Separate device activation from payment approval

Enrolling another device can form part of a scam. Liability depends on the recorded sequence and legal conditions.

Leggi di più →
Diritto e leggi

Profile scraping: Public visibility is not unrestricted reuse permission

Collecting and repurposing information are separate operations. Public visibility and technical accessibility do not settle every legal question.

Leggi di più →
Sentenze chiave

CJEU C-252/21: Data protection can matter to competition oversight

The Meta judgment connects abuse-of-dominance assessment with data protection while preserving institutional boundaries.

Leggi di più →
Sentenze chiaveAlta

BGH VI ZR 576/19: Access can include personal information in internal notes

Access is not limited to master records, but it is not unrestricted inspection of every internal document either.

Leggi di più →
Sentenze chiave

CJEU C-487/21: Data copies must be faithful and intelligible

A summary is not always enough. Context or document extracts may be necessary to understand personal data effectively.

Leggi di più →
Sentenze chiaveAlta

LAION and text-and-data mining: Interpreting a specific copyright dispute

An appellate decision is available. Research, dataset preparation and commercial model training are different activities.

Leggi di più →
Diritto e leggi

Payment-card data breaches: Record effects and assess claims carefully

Immediate protection and legal assessment are separate tasks. A court name does not replace identifiable evidence.

Leggi di più →
Diritto e leggi

Messengers and address books: Manage business contacts deliberately

Message encryption does not settle contact synchronisation. Data types, legal basis and device settings need separate examination.

Leggi di più →
Sentenze chiave

CJEU C-175/20: Tax-information requests must remain proportionate

The vehicle-advertising platform case does not impose a general ban on tax authorities requesting platform data.

Leggi di più →
Case di curaAltaIn evidenza

Cameras in care rooms: Consider images, audio and everyone affected

Well-intended monitoring can intrude deeply into privacy. Criminal liability and wider lawfulness are separate questions.

Leggi di più →
Case di curaAltaIn evidenza

Ransomware in care: Prepare availability and recovery

A backup is not a tested continuity process. Essential information needs an appropriate route back into use.

Leggi di più →
Case di curaAlta

Care sensors and GPS: Separate privacy from deprivation of liberty

Measuring, alerting and restricting movement are different functions. Assess the complete arrangement.

Leggi di più →
Assistenza ambulatorialeAltaIn evidenza

Messaging in care: Secure communication needs more than encrypted content

Patient information, contacts, devices and recipients need an agreed workflow.

Leggi di più →
Assistenza ambulatorialeAlta

Wound photographs and cloud backups: Follow the whole image lifecycle

Capture, gallery, synchronisation and the patient record are different stages requiring control.

Leggi di più →
Assistenza ambulatorialeAlta

A stolen work device: Assess, record and notify where required

Encryption, accessible keys, affected information and availability determine the risk, not device loss alone.

Leggi di più →
Assistenza ambulatorialeAltaIn evidenza

Care documentation: Traceable correction is different from falsification

Original entries and change timing must remain recognisable. Employment consequences require a separate case-specific assessment.

Leggi di più →
Sanitaria / ortopediaAltaIn evidenza

Medical supply businesses and TI: Plan connection and digital prescriptions separately

Connecting to the infrastructure does not mean every prescription workflow is already operational.

Leggi di più →
Sanitaria / ortopedia

3D body scans: Distinguish health information from biometric identification

A medical fitting scan can be sensitive without being the same operation as biometric access control.

Leggi di più →
Sanitaria / ortopedia

Section 302 SGB V billing: Organise content and transmission clearly

Billing may contain sensitive information. Data requirements and technical safeguards are related but separate matters.

Leggi di più →
PodologiaAlta

Podiatry and TI: Plan the actual connection pathway

Applications, credentials and legal deadlines are separate questions in infrastructure onboarding.

Leggi di più →
Podologia

Podiatry photographs: Separate treatment documentation from publication

Clinical records and marketing images serve different purposes. Required clinical documentation needs professional determination.

Leggi di più →
Podologia

Open screens and treatment records: Combine confidentiality with retention

Long-term availability does not remove the need for daily access control. There is no universal statutory two-minute screen rule here.

Leggi di più →
Consulenza assistenzialeAltaIn evidenza

Video care consultations under section 37 SGB XI: Check requirements, not brand slogans

Eligibility, first home visit, time limit and technical requirements must be considered together.

Leggi di più →
Consulenza assistenzialeAlta

Sending consultation records securely: Check content, transport and recipient

TLS protects transport; S/MIME and OpenPGP can protect message content. The agreed reporting procedure also matters.

Leggi di più →
Case di curaAlta

WeChat in care: Confidential workplace messaging needs an approved framework

A private group is not an approved care channel. Assess data, recipients, work devices and alternatives before sharing resident information.

Leggi di più →
Assistenza ambulatorialeAltaIn evidenza

TI Messenger and WhatsApp Business: Compare more than encryption

TI Messenger provides a governed healthcare messaging framework. Approval does not replace endpoint management, recipient checks or documentation.

Leggi di più →
Diritto e leggiAlta

Microsoft Teams in hospitals: Approve a defined use, not a product label

Administrative meetings and clinical case conferences need different assessments. Check contracts, access, recordings and the actual configuration.

Leggi di più →
Case di curaAlta

Signal and Threema in care: Encryption is only part of approval

Encrypted messages do not automatically solve account management, device loss or record keeping. Assess the whole workplace workflow.

Leggi di più →
Assistenza ambulatorialeAlta

Voice messages in clinical handovers: Control storage and acknowledgement

Audio may expose confidential information or be misunderstood. Define handover, documentation and lost-device procedures before relying on it.

Leggi di più →
Ricerca IAAltaIn evidenza

AI dictation at the bedside: Recording authority, data flow and clinical review

Dictating notes differs from recording a patient conversation. Assess recording authority, processing and human approval before introducing AI notes.

Leggi di più →
Minacce IAAlta

ChatGPT for clinical and care reports: Separate confidential inputs from plausible errors

Fluent AI writing is not evidence of a correct report. Approved processing and professional review matter more than a product label.

Leggi di più →
Diritto e leggiAlta

DeepL and Google Translate at the bedside: Set clinical and privacy boundaries

Language support must preserve understandable information and confidentiality. Distinguish routine assistance from high-stakes clinical communication.

Leggi di più →
Minacce IAAlta

AI medication advice: A plausible answer is not a validated prescription

Language models can produce convincing errors. Define boundaries, approved sources and professional verification for medication-related questions.

Leggi di più →
Diritto e leggiAlta

AI symptom checkers: Intended purpose matters more than the chatbot label

Appointment support and individual medical decision support are different uses. Define and verify the scope before putting an assistant on a website.

Leggi di più →
Assistenza ambulatorialeAlta

AI wound-analysis apps: Check intended purpose, image quality and storage

Storing a wound image differs from providing treatment recommendations. Assess intended purpose, correct patient attribution and the complete image path.

Leggi di più →
Diritto e leggiAltaIn evidenza

Healthcare cloud and section 393 SGB V: Assess scope and C5 evidence precisely

Section 393 sets specific cloud conditions and exceptions. A German server location or an arbitrary ISO certificate is not a complete assessment.

Leggi di più →
Diritto e leggiAlta

CLOUD Act and health data: Separate residency, authority requests and encryption

An EU data centre does not answer every access question. Assess the actual service, transfer arrangements and technical controls without blanket claims.

Leggi di più →
Assistenza ambulatorialeAltaIn evidenza

Care-sector cloud backups: Security depends on tested restoration

Synchronisation, version history and backups serve different purposes. Plan independent copies and verify that the care application can actually restart.

Leggi di più →
Case di curaAlta

Care-home SaaS outages: Plan continuity before the documentation service fails

Availability promises do not replace downtime procedures. Prepare critical information, fallback workflows and a controlled return to normal operation.

Leggi di più →
Minacce IAAlta

Cloud tenant isolation: Hiding other patients in the interface is not enough

Shared platforms need object-level authorisation and explicit isolation tests. Handle unexpected records without extending an exposure.

Leggi di più →
Case di curaAlta

Voice assistants in care rooms: Support independence while limiting capture

Voice control may help residents. Assess microphone behaviour, visitors, care conversations and residents’ choices before deployment.

Leggi di più →
Case di curaAlta

Radar fall detection: No camera image does not mean no personal data

Image-free sensing may reduce data collection. Still assess identification, alarm handling and the limits of the actual product.

Leggi di più →
Case di curaAlta

Wearables and dementia: Separate location data, autonomy and restrictive actions

A tracker is neither automatically deprivation of liberty nor automatically lawful care. Assess wishes, authority, data and actual intervention.

Leggi di più →
PodologiaAlta

Foot scans and 3D printing: Protect health information in the workshop too

Measurement and CAD files can identify patients. Manage attribution, manufacturing recipients and transfers throughout the production workflow.

Leggi di più →
Case di curaAlta

Care-home Wi-Fi: Segment networks by need, not a rigid four-VLAN recipe

Administrative systems, devices and guest networks need different access. Plan and test controlled communication without disrupting care.

Leggi di più →
Assistenza ambulatorialeAltaIn evidenza

The ePA in care: Authorised access, not a complete record at one click

The ePA can support care information exchange. Understand access rights, objections, missing documents and the facility’s own documentation responsibilities.

Leggi di più →
Assistenza ambulatorialeAlta

Video wound consultations: Assess the actual service and reimbursement framework

Video consultations do not all follow identical billing rules. Check the applicable agreement, provider evidence and clinical suitability.

Leggi di più →
Case di curaAlta

Emergency dashboards: Prioritise provenance and freshness over real-time promises

Interfaces can support emergency handover. FHIR alone does not guarantee access control, current information or correct patient matching.

Leggi di più →
Consulenza assistenzialeAlta

USB drives and private laptops in care advice: Control endpoints and transfers

Mobile care advice needs protected, usable records. Manage endpoints and incidents without claiming automatic fines or liability.

Leggi di più →
Sanitaria / ortopediaAlta

Medical-supply billing under section 302 SGB V: Secure channels, not format myths

Billing needs controlled interfaces and credentials. The name EDIFACT alone does not establish executable malware or a proven ransomware incident.

Leggi di più →