Sweet Security has unveiled significant advancements in its AI security offerings with the launch of Agentic AI Blocking. This feature allows organizations to counteract rogue AI behavior in real time, extending Sweet’s proactive runtime enforcement beyond the cloud to include intelligent agents operating alongside cloud resources. This is more significant than it looks, as the integration of AI in business processes has never been more pervasive.
The Growing AI Challenge
With about 80% of global businesses now integrating AI into their operations, the stakes are escalating. AI agents can access sensitive data independently and perform actions without direct oversight, which raises alarms about potential misconduct. Current security measures typically alert organizations after unauthorized actions have been executed, leaving vulnerabilities that can be exploited before any corrective action is taken. This reactive approach is inadequate in a landscape where AI capabilities are accelerating.
CyberNews Wire
Real-Time Intervention
“The dynamic nature of AI has reduced the cost of executing attacks, putting fully autonomous software into everyday business processes,” stated Dror Kashti, CEO and co-founder of Sweet Security. He emphasized the critical need for real-time action, detailing how Sweet’s system intervenes instantly: “If an agent attempts to access unauthorized data, Sweet intervenes immediately, providing clear feedback on the block. This prevents any adverse actions and ensures security remains intact.”
This direct, real-time action is unprecedented in the field. Many organizations might be reluctant to transition from passive monitoring to active interception, especially if they don't fully grasp the implications of AI's rapid proliferation.
Key Features of Agentic AI Blocking
- Real-Time Termination: Blocks unauthorized calls and sessions during their execution. This feature is critical for organizations aiming to protect data integrity in a world increasingly driven by automated decisions.
- Data Protection: Prevents sensitive information, including secrets and personally identifiable information (PII), from being accessed unlawfully, which helps maintain compliance with data protection regulations.
- Prompt Injection Defense: Stops harmful prompt injections live to direct agents toward appropriate actions, significantly reducing the risk of AI being manipulated.
Sweet's Learning Loop
What sets Sweet apart is its sophisticated runtime analysis capabilities, embodied in the Sweet Learning Loop. This system constantly assesses and learns from over one billion events daily. Such extensive data collection enables the platform to evaluate every application and agent against a critical benchmark: “execute exactly as intended, without deviation.” This level of precision permits enforcement without causing interruptions to live operations, ensuring business continuity. Interestingly, this is one area where many competitors falter—they can't match Sweet's ability to adapt on-the-fly without service disruption.
Industry Perspectives
Experts in the field increasingly recognize runtime enforcement as a non-negotiable aspect of AI security. Analysts are pushing the narrative that the most effective solutions will be those that preemptively thwart risky behaviors rather than simply issue alerts. If you’re working in this space, the message is clear: waiting for breaches to occur before notifying stakeholders isn’t enough anymore. Sweet is one of the few platforms that tackles both cloud and AI security challenges together in one cohesive system, enhancing protective measures across the board. As regulatory frameworks tighten, solutions like Sweet's will likely become not just preferable but essential.
Success Stories and Validation
Already, Sweet has demonstrated its capabilities on a large scale, notably with clients like Zoomd. They benefit from Sweet's protection across thousands of cloud applications, where any downtime could have serious market repercussions. This kind of performance is what organizations should expect from their security partners. It reflects how critical proactive security measures have become in today's business environment. “Sweet enabled us to embrace AI confidently throughout our operations,” shared Niv Sharoni, CTO at Zoomd. He highlighted the major advantage distinct to Sweet’s approach: “Most tools inform you of issues after they occur. Sweet acts on threats instantly, circumventing potential damage entirely. This is a pivotal shift from merely monitoring risks to actively neutralizing them.”
Future Outlook
The implications of Sweet Security's advancements cannot be overstated. As AI continues to evolve, the type of security provided by solutions like Agentic AI Blocking will likely redefine what organizations expect from their operational security frameworks. The future of security won’t just be about preventing breaches after they happen; it will involve anticipating malicious activities before they materialize. For those attending Black Hat USA 2026, there will be an in-person demonstration of Sweet's autonomous protective capabilities at Booth 5721. More details are available at https://hi.sweet.security/black-hat-2026.
About Sweet Security
Founded in 2023 by veterans of elite cyber units in the IDF and backed by significant investment totaling $120 million, Sweet Security specializes in preemptive runtime enforcement for both AI and cloud environments. While largely focused on detection and alerts, Sweet provides an alternative model by actively blocking malicious behaviors before they escalate. This is a departure from conventional thinking and reflects a growing awareness that the old models just don't work anymore. Visit www.sweet.security for more information.
Contact
Noa Glumcher
Sweet Security, CMO
