CSC News Table of Contents
Agentic AI Security moved into a new phase as Check Point said it will acquire Lakera, a fast-rising startup that builds defenses for large language models and AI agents. The agreement reflects how quickly enterprises need practical safeguards as AI agents automate tasks and make decisions.
The acquisition underscores a broader industry shift toward proactive controls for AI systems. You can read the original report here.
Agentic AI Security: Key Takeaway
- Agentic AI Security strengthens enterprise defenses by combining Lakera’s AI guardrails with Check Point’s threat intelligence and platform reach.
Why This Deal Matters Now
Check Point’s move brings together threat intelligence at scale with Lakera’s specialized controls for model misuse and prompt injection. That pairing is timely because Agentic AI Security must protect systems that gather context, call tools, and take actions on behalf of users.
These agents operate across email, cloud, and data stores, which increases exposure to jailbreaks, data leakage, and supply chain risks.
Enterprises are expanding AI pilots into production, and many are already assessing governance frameworks. Guidance from the NIST AI Risk Management Framework and the OWASP Top 10 for LLM Applications makes one point clear.
Agentic AI Security needs layered controls that block malicious inputs, enforce output policies, and monitor agent behavior in real time.
What Lakera Brings to the Table
Founded in Switzerland, Lakera is known for building guardrails that filter malicious prompts, detect policy violations, and prevent data exfiltration.
Its technology focuses on adversarial testing, robust content filters, and dynamic policies. That aligns closely with Agentic AI Security because model-safe behavior depends on reliable inputs, controlled tool use, and strict data boundaries.
Lakera also helped educate the industry by popularizing hands-on tests that expose prompt injection and jailbreak tricks, which have become essential pre-deployment checks for agent workflows.
Agentic AI Security benefits when guardrails are built into model endpoints and orchestration layers. If the combined company embeds Lakera controls inside Check Point’s product stack, customers could see unified policies that follow agents across email, chat, code, cloud, and network paths. That would reduce the complexity of deploying LLM apps at scale.
How Check Point Can Scale the Vision
Check Point has global reach, a mature partner ecosystem, and deep research fueled by threat telemetry. Integrating that data with Lakera policies could advance Agentic AI Security by detecting attack patterns faster and updating guardrails as new tactics emerge.
For example, the ability to correlate email-borne lures with model misuse attempts would help teams respond before an agent takes unsafe actions.
For organizations building their own agent platforms, this deal signals the need to invest in strong identity, hardened secrets, and encrypted storage.
Agentic AI Security improves when teams pair model guardrails with secure password management. Tools like 1Password or Passpack keep tokens and API keys out of code and away from unauthorized users.
Market Context and Risk Backdrop
Recent research and real-world incidents show why Agentic AI Security is top of mind. Security teams are tracking prompt injection campaigns and testing red-team methods that trick models into leaking secrets or executing unsafe tools.
For background on how prompt attacks evolve, see this explainer on prompt injection risks and this overview of a Microsoft prompt injection challenge designed to raise defensive awareness. Benchmark efforts, such as Open CyberSoC Eval, further highlight where defenses must mature.
Agentic AI Security also depends on basic cyber hygiene. Backups protect the data that agents touch, and services like IDrive make resilient backup plans easier for small teams.
Vulnerability management tools from Tenable help reduce the attack surface around AI pipelines. Network visibility from Auvik can reveal unusual traffic patterns when agents call tools, services, or third-party APIs.
Standards, Testing, and Continuous Controls
Agentic AI Security improves when teams align testing with known threat models. The MITRE ATLAS knowledge base documents adversary behavior against machine learning systems.
That mapping helps teams prioritize controls such as input validation, tool-use restrictions, and continuous monitoring with policy rollback. For regulated sectors, implementing data loss prevention and email authentication adds a crucial layer.
Services such as EasyDMARC can strengthen email trust, while Tresorit supports encrypted file workflows for sensitive prompts and outputs.
Agentic AI Security also touches privacy. If staff or executives test AI tools with personal data, a removal service like Optery can reduce exposure on data broker sites.
For a deeper look at privacy risks and mitigations, see our related coverage on personal info removal tools and an emerging AI data protection solution designed to protect training sets.
Implications for Enterprise Defenders
The upside is clear. Agentic AI Security can advance with unified policies that follow agents across endpoints, networks, and cloud apps. Check Point’s scale can turn Lakera’s guardrails into widely available controls with enterprise support, which removes friction for security leaders who need to move pilots into production without adding separate vendors.
Better alignment with standards and faster updates based on global telemetry could reduce the window of exposure to jailbreaks, prompt injections, and data leakage.
There are trade-offs to consider. Agentic AI Security works best when it is transparent and does not slow developers. Complex deployments can cause delays if teams must refactor agent code to use new guardrails.
Centralized controls may also raise questions about model latency, resource utilization, and false positives that block useful outputs. Buyers should evaluate how policies are tuned, how exceptions are reviewed, and how the platform logs agent actions for audit and forensics.
Enterprises should prepare with training and simulations. Programs like CyberUpgrade help upskill staff on secure AI practices. Building an internal learning path with LearnWorlds can support continuous education across developer and security teams.
For third-party testing, marketplaces such as GetTrusted can help source vetted cybersecurity services to stress-test agent pipelines. Agentic AI Security is not a one-time control but an ongoing program that adapts as models and attackers change.
Conclusion
Check Point’s agreement to buy Lakera is a meaningful signal. Agent workflows are expanding, and Agentic AI Security needs to be part of the core security stack, not an afterthought. Pairing guardrails with threat intelligence, identity controls, and resilient infrastructure gives teams a better chance to stay ahead.
As the market evolves, keep an eye on how these capabilities integrate into everyday products. Agentic AI Security will succeed when safer defaults make it easy for developers to build, for security teams to monitor, and for business leaders to trust AI outcomes.
FAQs
What is Agentic AI Security?
– A security approach that protects AI agents as they interpret prompts, call tools, access data, and take actions on behalf of users.
Why is this acquisition important?
– It combines model guardrails with enterprise threat intelligence, making it easier to deploy safer AI agents at scale.
How do guardrails help?
– They filter malicious prompts, enforce content policies, and prevent unsafe tool use or data exfiltration.
What should teams do first?
– Map agent tasks, secure identities and secrets, and align controls with frameworks like NIST AI RMF and OWASP LLM guidance.
Can small teams improve quickly?
– Yes, by adopting password managers, secure storage, reliable backups, and tested guardrails to reduce common risks.
Where do backups fit in?
– Backups protect the data agents rely on, enabling recovery after errors, corruption, or ransomware.
Do we need specialized training?
– Training reduces mistakes and prepares teams to recognize and mitigate prompt injection and jailbreak tactics.
About Check Point Software Technologies
Check Point Software Technologies is a global cybersecurity vendor known for network security, cloud protection, and endpoint defenses. The company serves enterprises, service providers, and public sector organizations with integrated products and threat intelligence.
Its portfolio spans firewalls, email security, mobile protection, and cloud-native controls, all backed by research teams that track adversary behavior around the world. With the rise of AI-driven tools, Check Point is extending its platform to support Agentic AI Security so customers can deploy modern applications with confidence and control.
About Gil Shwed
Gil Shwed is the founder of Check Point Software Technologies and a long-time cybersecurity innovator. He helped pioneer stateful inspection technology, which set a foundation for modern network security and influenced how enterprises protect critical systems.
Under his leadership, Check Point grew into a global provider with a strong focus on research and integrated defenses.
As AI becomes central to business operations, his vision emphasizes practical controls and trusted platforms, priorities that align closely with the goals of Agentic AI Security.
Additional Resources and Related Reading
To deepen your understanding of risks and defenses, consider these resources on prompt attacks and benchmarks, including how benchmarks from industry leaders inform real-world testing.
For incident readiness shaped by AI-era threats, explore incident response strategies. If your team needs secure collaboration around model outputs, encrypted cloud storage from Tresorit and DMARC enforcement via EasyDMARC can help reduce data leakage and phishing risks.
For vulnerability scanning around AI components, visit Tenable’s solutions. When agents handle operational workflows, dependable feedback management via Zonka Feedback can capture issues quickly so teams tune policies and improve outcomes.
