Cyber Brief: AI safety, ransomware tactics, charity breach

Today's brief reflects the growing operational complexity organisations face as AI capabilities advance, attack methods evolve, and supply chain dependencies create unexpected exposure. OpenAI's decision to pause internal work on its next AI model after discovering significant cybersecurity capabilities highlights how rapidly AI tools are developing beyond current safety frameworks. Meanwhile, ransomware groups are refining their social engineering tactics by targeting mid-level IT managers rather than senior executives, and a breach at a UK charity software provider has exposed sensitive data across healthcare and victim support organisations.

OpenAI pauses next AI model after internal tests reveal advanced cyber capabilities

The Hacker News reports that OpenAI has paused some internal activities involving its upcoming AI model, codenamed Astra, after internal evaluations found it had made significant advancements in agentic coding and cybersecurity capabilities. The company stated it is implementing additional security controls for higher-capability models and associated activities, including isolated testing environments. The decision follows OpenAI's own safety protocols, which require heightened scrutiny when models demonstrate capabilities that could be used for offensive cyber operations or autonomous exploitation of vulnerabilities.

This development matters because it demonstrates that AI models are now reaching capability thresholds that even their developers consider operationally risky to deploy without additional safeguards. For UK organisations already integrating AI tools into development workflows, security operations and business processes, this is a reminder that the technology is advancing faster than many existing governance frameworks can accommodate. The fact that a leading AI company has chosen to pause work on a model due to cyber risk concerns suggests that organisations should be reviewing how they assess, approve and monitor AI tools used internally, particularly those with code generation, automation or system access capabilities.

Why it matters

For UK businesses, this is a prompt to review how AI tools are being introduced, approved and monitored across the organisation. Consider whether there is clear ownership of AI risk assessment, whether development and security teams understand the capabilities of tools already in use, and whether there are controls in place to prevent AI systems from being used in ways that could introduce security risk or bypass existing approval processes.

Source: The Hacker News

Ransomware groups increasingly target mid-level IT managers rather than senior executives

The Register reports that ransomware gangs are shifting their social engineering tactics away from targeting CEOs and senior executives, instead focusing on IT managers, typically in their 40s, who have privileged access to systems but may be less prepared for sophisticated phishing or vishing attacks. The article explains that these mid-level technical staff often have administrative credentials, access to backup systems and the ability to disable security controls, making them valuable targets. Attackers are using increasingly convincing pretexts, including fake IT support calls, vendor impersonation and urgent requests that exploit the pressure and workload these individuals face.

This shift in attacker behaviour reflects a more operationally aware approach from ransomware groups, who understand that technical staff with day-to-day system access are often easier to manipulate than board-level executives who are more likely to have received targeted security awareness training. For UK organisations, this highlights a gap in how security awareness and incident response preparedness is distributed across the business. IT managers are often expected to be the first line of defence, yet they may not receive the same level of support, training or scrutiny as senior leadership when it comes to recognising and responding to social engineering attempts.

Why it matters

For many organisations, this is a prompt to review whether security awareness training, phishing simulation and incident response guidance is reaching technical and operational staff who hold privileged access. Consider whether IT managers know how to verify unexpected requests, whether they feel able to challenge urgent demands without fear of blame, and whether there are clear escalation routes when something feels wrong.

Source: The Register

UK healthcare and victim support charities affected by Beacon CRM cyber incident

Infosecurity Magazine reports that Beacon, a widely used customer relationship management (CRM) platform for charities, has informed approximately 1,500 customer organisations that its databases were accessed by an unauthorised actor, with data likely exfiltrated. Beacon provides CRM and fundraising software to charities across the UK, including organisations in the healthcare and victim support sectors. The breach means that sensitive donor information, beneficiary records and operational data held by these charities may have been compromised. Beacon has stated it is working with cybersecurity specialists and law enforcement, and has informed affected customers so they can assess their own exposure and notify individuals where required.

This incident highlights the operational and reputational risk that comes from relying on third-party platforms to manage sensitive data, particularly in sectors where trust and confidentiality are fundamental to the relationship between organisations and the people they serve. For UK charities, healthcare providers and other organisations using Beacon or similar platforms, this is a reminder that supply chain risk is not just a technical concern but a governance and safeguarding issue. The breach also underscores the importance of understanding what data is held by third parties, how it is protected, and what notification and response obligations exist when things go wrong.

Why it matters

For UK businesses and charities, this is a prompt to review third-party risk management processes, particularly for platforms that hold sensitive personal data on behalf of the organisation. Consider whether contracts clearly define data protection responsibilities, whether there are processes in place to monitor supplier security posture, and whether incident response plans account for breaches that originate outside the organisation's direct control.

Source: Infosecurity Magazine

Critical vulnerability in Atlassian Rovo AI assistant could have exposed Jira and Confluence data

SecurityWeek reports that security researchers at Varonis discovered a critical vulnerability in Atlassian's Rovo AI assistant that could have been exploited to steal data from Confluence, Jira and SharePoint. The attack method, dubbed RovoBlast, involved tricking the AI assistant into following attacker-controlled instructions embedded in content it was reading, causing it to collect sensitive data accessible to a signed-in user and send it to an external server. The vulnerability required only a single click from a user interacting with the AI assistant. Atlassian has since addressed the issue, though researchers noted that one of the attack vectors they identified independently may still be open.

This vulnerability is significant because it demonstrates how AI assistants, which are increasingly being integrated into enterprise collaboration platforms, can be manipulated to bypass traditional access controls and exfiltrate data in ways that would not be possible through conventional attack methods. For UK organisations using Atlassian products or similar AI-enhanced tools, this is a reminder that AI features introduce new attack surfaces that may not be covered by existing security controls, monitoring or user awareness training. The fact that the vulnerability could be triggered by a single user action, without requiring elevated privileges or complex exploitation, makes it particularly concerning from an operational risk perspective.

Why it matters

For UK businesses, this is a prompt to review how AI-enhanced features are being enabled across collaboration and productivity platforms. Consider whether security teams understand the data access permissions these AI tools inherit, whether there are controls in place to limit what data AI assistants can read or share, and whether users are aware of the risks associated with interacting with AI-generated content or prompts.

Source: SecurityWeek

Today's Key Actions

• Review how AI tools are being introduced, approved and monitored across the organisation, particularly those with code generation, automation or system access capabilities, and ensure there is clear ownership of AI risk assessment.
• Ensure security awareness training, phishing simulation and incident response guidance reaches technical and operational staff who hold privileged access, and that IT managers know how to verify unexpected requests and escalate concerns without fear of blame.
• Review third-party risk management processes for platforms that hold sensitive personal data, ensuring contracts clearly define data protection responsibilities and incident response plans account for breaches originating outside direct organisational control.
• Review how AI-enhanced features are enabled across collaboration platforms, ensuring security teams understand data access permissions these tools inherit and users are aware of risks associated with AI-generated content.
• Ensure there is clear ownership across the organisation for reviewing and acting on emerging risks in AI adoption, supply chain security, social engineering resilience and third-party data protection.

Secarma Insight

Today's stories reflect a common thread: security risk increasingly emerges from the intersection of new technology, trusted relationships and operational complexity. Whether it is AI capabilities advancing faster than safety frameworks, attackers exploiting the trust placed in mid-level staff, or supply chain dependencies creating unexpected exposure, the challenge is not just technical but organisational. Mature security practice comes from understanding where these intersections exist, ensuring the right people have the right context to make good decisions, and building habits of review, escalation and accountability that are already in place before incidents happen. The organisations that manage these risks well are those that treat security as a shared responsibility, supported by clear ownership, practical guidance and a culture where asking questions is valued over appearing certain.