A Fireside Panel on Red Team, Blue Team & Purple Team AI Security
Introduction
Artificial intelligence is moving rapidly from experimentation into the heart of enterprise operations. AI is now being integrated into customer platforms, software development, cybersecurity, business processes, analytics, decision-making, and critical workflows.
But as organisations accelerate AI adoption, a fundamental question is becoming impossible to ignore:
How do you secure AI when the technology itself introduces new attack surfaces, behaviours, and risks?
AI security requires a different mindset. Traditional cybersecurity controls remain essential, but organisations must also understand threats such as prompt manipulation, adversarial attacks, data leakage, model abuse, insecure integrations, excessive AI permissions, and attacks against the infrastructure surrounding AI systems.
At the same time, organisations need to know whether their defensive controls actually work when AI systems are under pressure.
This is where offensive and defensive security perspectives become critical.
The Fireside Panel: Securing AI in Action at Next IT Security | Stockholm 2026 will bring together Red Team, Blue Team, and Purple Team perspectives to examine how organisations can identify AI vulnerabilities, defend AI operations, and continuously improve their security posture.
Rather than treating AI security as a theoretical challenge, the discussion will focus on the practical realities of securing AI in enterprise environments.
Why AI Security Matters Now
The speed of AI adoption is creating an equally rapid expansion of the AI attack surface.
Organisations are connecting AI models to corporate data, APIs, cloud infrastructure, business applications, plugins, agents, and automated workflows. These connections can create significant value, but they can also introduce pathways that attackers may attempt to exploit.
A vulnerability in an AI system may not look like a traditional vulnerability. An attacker might manipulate an AI application’s instructions, extract sensitive information through carefully constructed interactions, exploit an insecure integration, abuse excessive permissions, or influence an automated process into performing an unintended action.
For security leaders, this creates a new challenge: AI must be tested and defended as an operational system, not simply evaluated as a model.
That requires visibility across the entire AI ecosystem.
The Securing AI in Action panel will explore how offensive testing, defensive operations, and cross-team collaboration can help organisations understand and manage these emerging risks.
What Attendees Will Explore
The discussion will focus on:
- The emerging attack surface surrounding enterprise AI.
- How attackers can target AI models, applications, data, and integrations.
- How security teams can identify AI-specific vulnerabilities.
- How organisations can monitor and defend AI operations.
- How red and blue teams can collaborate more effectively.
- How AI security findings can strengthen governance and risk management.
- How organisations can build resilience as AI adoption scales.
Red Team Perspective: Identifying AI Vulnerabilities
Think Like an Attacker Before the Attacker Does
Red teams have always played a critical role in cybersecurity by asking a simple but powerful question:
βHow could this system be attacked?β
In the world of AI, that question becomes considerably more complex.
AI systems can behave differently depending on the inputs they receive, the context provided to them, the data they can access, and the tools or systems they are connected to. This creates opportunities for attackers to search for unexpected pathways through an AI environment.
AI red teaming allows organisations to simulate these scenarios in a controlled and authorised manner.
Offensive security specialists can test whether AI applications can be manipulated, whether safeguards can be bypassed, whether sensitive information can be exposed, and whether AI systems can be persuaded to perform actions outside their intended purpose.
But effective AI red teaming goes beyond trying to βbreak the model.β
It involves examining the entire AI attack surface β including models, applications, APIs, identities, data sources, integrations, infrastructure, and human interaction.
Impact on Next-Generation IT Security
AI red teams can help organisations:
- Discover previously unknown AI vulnerabilities.
- Test resistance to adversarial manipulation.
- Identify weaknesses in AI application security.
- Assess exposure of sensitive information.
- Test AI integrations, APIs, plugins, and connected tools.
- Identify excessive permissions and unintended capabilities.
- Challenge assumptions about AI safeguards.
- Evaluate potentially unsafe or biased behaviour.
At Next IT Security | Stockholm 2026, the Red Team perspective will demonstrate why organisations need to understand AI from an adversary’s point of view before real attackers discover the weaknesses first.
Testing AI Without Creating New Risks
Offensive Security Needs Boundaries
Realistic testing is essential, but AI security assessments must be conducted responsibly.
Enterprise AI systems may have access to confidential business information, customer data, internal applications, development environments, or systems capable of taking automated actions. Uncontrolled testing could therefore create operational, legal, privacy, or security risks of its own.
Effective AI red teaming requires clearly defined rules of engagement.
Security teams must establish what can be tested, which data can be accessed, what actions are permitted, and how potentially sensitive findings will be handled.
Ethical considerations are equally important.
AI security testing may reveal unexpected bias, unsafe behaviour, or weaknesses that could affect customers or employees. Responsible testing therefore requires organisations to balance realistic attack simulation with appropriate safeguards.
The objective is not simply to prove that an AI system can be broken.
The objective is to understand how it could fail β and make it stronger as a result.
Why This Matters
For security leaders, responsible AI testing can provide:
- Better visibility into real-world AI risks.
- Stronger evidence for AI risk assessments.
- More informed security investment decisions.
- Clearer remediation priorities.
- Greater confidence in AI deployments.
- A stronger foundation for responsible AI governance.
The Securing AI in Action panel will explore how organisations can push AI systems to their limits while maintaining the ethical, operational, and regulatory discipline expected in modern enterprise security.
Blue Team Perspective: Defending AI Operations
From Detecting Attacks to Protecting AI in Production
Finding vulnerabilities is only half the security challenge.
Once AI systems are deployed into production, organisations need to know what is happening inside them.
Blue teams are responsible for building that defensive layer.
For AI environments, this means monitoring not only the underlying infrastructure but also interactions with models, applications, APIs, identities, data sources, and connected tools.
Security teams need to establish what normal AI activity looks like and identify behaviour that could indicate misuse or compromise.
This can include unusual access patterns, suspicious requests, unexpected data flows, abnormal interactions, compromised credentials, or attempts to manipulate an AI application’s intended behaviour.
AI incident response must also evolve.
A security incident involving AI may not resemble a conventional malware attack. It could involve manipulated outputs, sensitive data exposure, compromised AI credentials, misuse of an AI agent, or an AI-enabled application being abused as a pathway into other systems.
Defenders therefore need visibility, context, and response procedures designed for AI-enabled environments.
Impact on Next-Generation IT Security
Blue teams are increasingly focused on:
- Continuous AI security monitoring.
- Identity and access management for AI systems.
- Protection of sensitive AI data.
- Detection of anomalous AI behaviour.
- AI-specific incident-response procedures.
- Secure configuration of models and AI applications.
- Protection of AI infrastructure and integrations.
- Maintaining operational continuity during security incidents.
At Next IT Security | Stockholm 2026, the Blue Team perspective will explore how organisations can defend AI in production while maintaining reliability, availability, and stakeholder confidence.
Defending AI Without Stopping Innovation
Security Must Enable Responsible Adoption
One of the biggest challenges facing security leaders is finding the balance between protection and innovation.
AI adoption cannot realistically be stopped simply because new risks exist. Organisations need to understand those risks and establish controls that allow AI to be used safely.
This means security teams need to move beyond a purely restrictive approach.
Effective AI security can involve appropriate access controls, monitoring, human oversight, segmentation, data protection, testing, incident-response capabilities, and clearly defined boundaries for autonomous actions.
Resilience is equally important.
Organisations should consider what happens if an AI service becomes unavailable, compromised, manipulated, or unreliable. Critical processes may need fallback procedures, human intervention, or the ability to isolate specific AI components without taking an entire business operation offline.
The Security Objective
The goal is not to eliminate every AI risk.
The goal is to understand, manage, monitor, and continuously reduce risk while allowing the organisation to benefit from AI.
That balance will be a central theme of the Securing AI in Action discussion.
Purple Team Perspective: Integrating Offensive and Defensive Insights
Turning Security Testing into Continuous Improvement
Red teams find weaknesses.
Blue teams defend against them.
Purple teams bring the two perspectives together.
This collaboration becomes particularly powerful in AI security because identifying a vulnerability is only part of the problem. Organisations also need to know whether they can detect exploitation, contain the activity, respond effectively, and prevent the same weakness from recurring.
Imagine a red team discovers that an AI application can be manipulated through a particular attack technique.
The next question is:
Would the blue team detect it?
If the answer is no, the organisation has identified not only a vulnerability but also a detection gap.
The teams can then work together to improve monitoring, update controls, strengthen response procedures, and test the environment again.
This creates a continuous feedback loop:
Attack β Detect β Respond β Improve β Retest
That cycle is at the heart of a mature AI security programme.
Impact on Next-Generation IT Security
Purple teaming can help organisations:
- Connect offensive findings with defensive improvements.
- Validate whether AI attacks can actually be detected.
- Improve AI incident-response capabilities.
- Strengthen security controls through realistic testing.
- Establish measurable security improvements.
- Build continuous feedback between security teams.
- Adapt controls as AI technologies and threats evolve.
At Next IT Security | Stockholm 2026, the Purple Team perspective will explore how collaboration can transform AI security from a periodic testing exercise into an ongoing resilience programme.
From AI Vulnerabilities to AI Governance
Security Findings Should Drive Better Decisions
AI security cannot exist in isolation from AI governance.
As organisations deploy AI across departments and business processes, leaders need to understand which systems are being used, what data they access, what decisions they influence, and what risks they introduce.
Security testing can provide valuable evidence for those governance processes.
A red team finding may reveal that an AI application has excessive access to sensitive information. A blue team may discover that the activity is difficult to monitor. A purple team exercise can then validate whether new controls actually address the problem.
This creates a much stronger approach to AI risk management.
Instead of relying entirely on policies, documentation, and assumptions, organisations can use real-world testing and operational evidence to understand whether their controls work.
Building a Stronger AI Security Framework
Organisations should consider:
- Clear ownership of AI security risks.
- Continuous testing throughout the AI lifecycle.
- Monitoring of AI systems in production.
- Documented security and governance controls.
- Defined accountability for AI-enabled decisions.
- Regular reassessment as AI capabilities change.
- Collaboration between security, risk, compliance, technology, and business teams.
The panel will examine how security teams can become an active part of AI governance rather than being brought into the conversation only after problems occur.
Protecting Trust in AI
Security Is a Foundation for Responsible AI
AI adoption depends heavily on trust.
Employees need confidence that AI tools will not expose sensitive information. Customers need confidence that automated systems are secure and reliable. Business leaders need confidence that AI-enabled processes will behave within acceptable boundaries.
Security is central to that confidence.
An AI system that produces impressive results but can easily be manipulated, abused, or compromised presents a significant organisational risk.
Trust therefore requires more than good model performance.
Organisations need to know:
- Can the system be manipulated?
- Can sensitive information be extracted?
- Can unauthorised users influence its behaviour?
- Can suspicious activity be detected?
- Can the organisation respond when something goes wrong?
- Can humans intervene when automated decisions become unreliable?
Red Team, Blue Team, and Purple Team practices can help answer these questions through continuous testing and validation.
Why Attendees Should Pay Attention
For CISOs, security leaders, AI leaders, risk professionals, architects, and technology decision-makers, AI security is rapidly becoming a strategic issue.
Understanding how offensive and defensive teams approach AI can help organisations make better decisions about where to invest, what to monitor, and how to prepare for emerging threats.
Preparing for AI Security Incidents Before They Happen
Resilience Starts Before Deployment
The best time to prepare for an AI security incident is before one happens.
Organisations should not wait until an AI system has been compromised, manipulated, or exposed to determine who is responsible for responding.
AI security exercises can help organisations test their assumptions in advance.
A realistic exercise might ask:
- What happens if an AI application starts leaking sensitive information?
- What if an AI agent attempts an unauthorised action?
- How would security teams detect manipulation?
- Who has authority to suspend the system?
- What happens to business operations if the AI service is unavailable?
- How quickly can the organisation investigate the incident?
- How are customers, employees, regulators, or other stakeholders informed?
These scenarios help expose gaps before they become real incidents.
Impact on Next-Generation IT Security
Organisations can strengthen AI incident readiness by:
- Including AI scenarios in security exercises.
- Establishing clear ownership and escalation procedures.
- Testing AI-specific detection capabilities.
- Developing fallback procedures for critical AI workflows.
- Defining when human intervention is required.
- Testing recovery and containment processes.
- Connecting AI incident response with broader crisis-management plans.
At Next IT Security | Stockholm 2026, the discussion will highlight why preparation, testing, and collaboration are essential for organisations deploying AI into increasingly critical environments.
Building an AI Security Culture
AI Security Is Everyone’s Responsibility
AI security cannot be owned exclusively by the cybersecurity team.
Developers build AI-enabled applications. Data teams manage information. Business units define use cases. Executives make investment decisions. Risk and compliance teams establish governance requirements. Security teams assess and defend the environment.
Each group sees a different part of the risk.
That is why collaboration is so important.
Red and blue teams provide complementary technical perspectives, while purple teaming creates a mechanism for bringing those insights together. The same principle can extend across the wider organisation.
When security becomes part of the AI development and deployment lifecycle, organisations are better positioned to identify risks early rather than attempting to fix them after deployment.
Building Security Into AI Adoption
A mature AI security culture encourages organisations to:
- Ask security questions before deploying new AI capabilities.
- Test AI systems continuously rather than once.
- Share security findings across teams.
- Make risk ownership explicit.
- Encourage responsible experimentation.
- Treat security as an enabler of trustworthy AI adoption.
The Securing AI in Action panel will explore what this collaborative security culture can look like in practice.
Why the Red, Blue & Purple Team Model Matters
AI security is too complex to be addressed from a single perspective.
The Red Team asks:
βHow could we attack this?β
The Blue Team asks:
βHow do we detect, defend, and respond?β
The Purple Team asks:
βHow do we connect what we learned and make the organisation stronger?β
Together, these perspectives create a continuous security cycle.
This approach moves AI security away from static compliance and toward active resilience.
It also creates a practical way for organisations to measure whether their security investments are delivering results.
A control is more meaningful when it has been tested.
A detection capability is more valuable when it has been validated.
A response plan is more credible when it has been exercised.
And an AI governance framework becomes more effective when it is informed by real-world security evidence.
What Attendees Can Take Away
The Fireside Panel: Securing AI in Action is designed to provide practical perspectives that organisations can take back to their own AI and cybersecurity programmes.
Attendees will gain insight into:
- How red teams approach AI vulnerabilities.
- How blue teams monitor and defend AI operations.
- How purple teams connect offensive and defensive security.
- How AI attack surfaces differ from traditional applications.
- How to test AI responsibly and within defined boundaries.
- How to improve AI incident-response readiness.
- How security findings can inform AI governance.
- How organisations can build continuous AI security validation.
- How to strengthen trust in AI-enabled business processes.
For organisations moving from AI experimentation toward large-scale deployment, these perspectives can provide a valuable framework for thinking about security before risks become incidents.
Looking Ahead
AI will continue to transform how organisations operate.
The question is no longer whether organisations will use AI, but whether they can scale it securely, responsibly, and resiliently.
That requires a shift in mindset.
Organisations must actively challenge their AI systems through offensive testing. They must build the monitoring, controls, and response capabilities required to defend those systems in production. And they must create continuous feedback loops that allow security teams to learn from every test, incident, and emerging threat.
The Fireside Panel: Securing AI in Action at Next IT Security | Stockholm 2026 will bring together the Red Team, Blue Team, and Purple Team perspectives to explore exactly how organisations can achieve this.
From identifying vulnerabilities to defending operations, from testing controls to strengthening governance, the discussion will focus on the practical steps required to build confidence in AI.
Because securing AI is not about assuming that technology will always behave as expected.
It is about being prepared when it doesn’t.
By combining offensive insight, defensive capability, and continuous collaboration, organisations can build AI environments that are not only innovative, but also secure, resilient, and worthy of trust.
Join the Conversation
The future of AI security will be shaped by organisations willing to test their assumptions, challenge their systems, and learn continuously.
The Securing AI in Action fireside panel at Next IT Security | Stockholm 2026 offers an opportunity to explore these challenges from multiple security perspectives and discover how Red, Blue, and Purple Team approaches can work together to secure AI in the real world.
Don’t just deploy AI. Secure it. Test it. Defend it. Learn from it.
