Sweet Security is making significant strides in AI security by unveiling its new Agentic AI Blocking feature. This development allows companies to proactively prevent unauthorized actions by AI agents in real time, extending Sweet's protective framework. The move reflects an urgent response to emerging vulnerabilities in a landscape increasingly dominated by AI technologies.
The Rise of AI in Business Operations
With AI becoming integral to enterprise operations, it’s alarming that about 80% of businesses are now utilizing AI agents that can interact with sensitive data autonomously. These agents can optimize processes and drive efficiencies, but they also pose a considerable risk. Traditional security tools primarily focus on detecting and alerting after the fact, leaving organizations vulnerable to unintended consequences long before they can respond adequately. This reactive approach creates a security gap that proactive measures aim to fill.
CyberNews Wire
Addressing Vulnerabilities in AI
According to Dror Kashti, CEO and co-founder of Sweet Security, “AI has collapsed the cost of attack. Someone has to decide, in the moment, what an agent is allowed to do.” This point emphasizes a critical vulnerability—without stringent oversight, AI systems can be exploited, leading to unauthorized access or misuse of sensitive information. Sweet Security’s approach directly addresses this by actively stopping unauthorized actions before they impact the organization, ensuring that nothing harmful occurs while also providing detailed reports on each incident. This is more significant than it looks; it aligns with a growing trend of organizations prioritizing real-time intervention over traditional post-incident analysis.
Key Features of Agentic AI Blocking
Key features of the new Agentic AI Blocking include:
- Real-time termination of unauthorized tool calls and sessions, allowing for immediate prevention of harmful actions at runtime.
- Protection of sensitive data from being accessed or transferred improperly, ensuring compliance and protecting corporate reputation.
- Immediate blockage of prompt injections that could redirect an agent's activities, a vulnerability often exploited in AI interactions.
The backbone of Sweet Security's platform lies in its confident runtime knowledge. The Sweet Learning Loop continuously processes over a billion runtime events daily, enabling it to understand the specific roles and permissions of each application and agent. This thorough understanding allows Sweet to enforce compliance without disrupting production environments, which is often a major concern for businesses wary of the implementation of new security measures.
Analyst Perspectives on Runtime Inspection
Analysts are increasingly aligning on the necessity of runtime inspection as a foundational element for securing AI agents. As AI continues to penetrate critical business processes, the expectation is that successful security products will emphasize preventive measures rather than simply alerting organizations to risks after they occur. This shift represents a pivot in how security is approached, especially in technology domains where speed and agility are paramount.
Sweet Security's unique offering integrates enforcement across both AI tools and cloud environments, aligning with current market trends that call for a more cohesive security strategy. This paradigm shift not only enhances real-time capabilities but also reflects a broader understanding of the risks inherent in autonomous systems.
Effectiveness and Adoption
Demonstrating the effectiveness of its solutions, Sweet already protects tens of thousands of applications at major companies like Zoomd. “Sweet gave us the confidence to adopt AI across the business,” remarked Niv Sharoni, CTO of Zoomd. This statement underlines how Sweet’s proactive blocking stands out from conventional monitoring tools that often report harmful behavior only after it occurs. It’s a sentiment echoed by others in the industry who recognize that security cannot afford to be an afterthought, especially when engaging AI technologies that operate on the fringes of established protocols.
Future Implications and Significance
If you're working in this space, consider the implications of Sweet Security's approach. As the dual threats of AI misuse and data breaches loom larger, companies will increasingly seek out solutions that offer both protection and insight into their operations. This trend towards proactive security enforcement will likely shape market dynamics, leading to tighter integration of security measures during the development phase of AI applications.
For those interested in witnessing Sweet Security’s advanced AI protection in action, the company will present live demonstrations at Black Hat USA 2026, Booth 5721. More information can be found at https://hi.sweet.security/black-hat-2026.
About Sweet Security
Founded in 2023 by specialists from elite cyber units in the IDF, Sweet Security focuses on providing proactive runtime protection for both cloud and AI technologies. Their method emphasizes blocking threats in real time, setting a new standard in enterprise security. Backed by substantial investment from notable venture firms, Sweet aims to safeguard the world’s leading companies. More details are available at www.sweet.security.
Contact Information
Noa Glumcher
Sweet Security, CMO
