Agentic AI and the New Attack Surface
Introduction
Agentic AI systems are autonomous systems that can plan, make decisions, and take actions by interacting with external tools and environments. But they shift the nature of risk by blurring the lines among input, decision, and execution. A conventional model generates an output and stops. An agent takes input, makes plans, invokes tools, updates its state and repeats the cycle. This creates a system where decisions are continuously revised through interaction with external tools and environments, rather than being fixed at the point of input.
This means the attack surface expands in size and becomes more dynamic. Instead of remaining confined to components as in traditional computational systems, they spread in layers and can continue to grow through time. To understand this shift, the system can be analysed through functional layers such as inputs, memory, reasoning, and execution, while recognising that risk does not remain isolated within these layers but emerges through their interaction.
Agentic AI Attack Surface
A layered view of how risks emerge across input, memory, reasoning, execution, and system integration, including feedback loops and cross-system dependencies that amplify vulnerabilities.
Input Layer: Where Untrusted Data Becomes Control
The entry point of an agent is no longer one prompt. The documents, APIs, files, system logs and the outputs of other agents can now be considered input. This diversity is significant due to the fact that every source of input carries its own trust assumptions, and in the majority of cases, they are weak.
The most obvious threat is prompt injection, where inputs are treated as instructions rather than data. Since inputs are treated as instructions, a virus, a malicious webpage, or a document can contain instructions that override system goals without necessarily being detected as something harmful.
Indirect prompt injection extends this risk beyond direct user interaction. Instead of targeting the interface, attackers compromise the retrieval process by embedding malicious instructions within external data sources. When the agent retrieves and processes the data, it treats the embedded content as legitimate input. As a result, the attack is executed through normal reasoning processes, allowing the system to act on untrusted data without recognising the manipulation.
Data poisoning also occurs at runtime. In contrast to classical poisoning (where training data is manipulated), runtime poisoning distorts the agent’s perception of its environment as it runs. This can change decisions without causing apparent failures.
Obfuscation introduces another indirect attacker vector. Encoded instructions or complicated forms may bypass human review but remain readable to the model. This creates asymmetry whereby the system knows more about the attack than those operating it. Once compromised at this layer, the agent implements compromised instructions which affect downstream operations.
Context and Memory: Persistence of Influence
Agentic systems depend on memory to operate efficiently. They often retain context across sessions and frequently store information between sessions.
This introduces a different type of risk: persistence. Through memory poisoning, attackers can insert false or adversarial information into sorted context, which then influences future decisions. Unlike prompt injection, which is often limited to a single interaction, this effect carries forward. Over time, the agent begins to operate on a distorted internal state, shaping decisions in ways that may not be immediately visible.
Another issue is cross-session leakage. Information in a particular context may be replayed in a different context when memory is being shared or there is insufficient memory separation. This is specifically dangerous in those systems that combine retrieval and long-term storage. The context management in itself becomes a weakness. Agents are required to make decisions on what to retain and what to discard. This is susceptible to attackers who can flood the context or manipulate what is still visible and indirectly affect reasoning.
The underlying problem is structural. Memory turns data into a state. Once state is corrupted, the system cannot easily distinguish valid knowledge from adversarial influence.
The issue is structural. Memory converts temporary data into a persistent state. Once this state is weakened, the system cannot reliably separate valid information from adversarial influence, making recovery significantly more difficult.
Reasoning and Planning: Manipulating Intent Without Breaking Logic
The reasoning layer is where agentic AI stands apart from traditional systems. The model no longer reacts to inputs alone. It actively breaks down objectives, analyses alternatives, and ranks actions.
At the reasoning stage, the nature of risk shifts. The concern is no longer limited to injecting instructions, but to influencing how decisions are made. One example is goal manipulation, where the agent subtly reinterprets its objective and produces outcomes that are technically correct but strategically harmful. Reasoning hijacking operates within intermediate steps, altering how constraints are evaluated or how trade-offs are prioritised. The system may remain internally consistent, which makes such deviations difficult to detect.
Tool selection becomes a critical control point. Agents decide which tools to use and when, so influencing these choices can redirect execution without directly accessing the tools themselves. Hallucinations also take on a different role here. In static systems, they remain errors. In agentic systems, they can trigger actions. A perceived need or incorrect judgement can translate into real-world consequences.
This layer introduces probabilistic failure. The system is not fully weakened, but it is nudged towards decisions that appear reasonable yet are incorrect. The risk lies in how those decisions are justified.
Tool and Execution: When Decisions Gain Reach
Once an agent begins interacting with tools, its behaviour extends beyond the model into external systems. APIs, databases, and services become part of the execution path.
One key risk is the use of unauthorised tools. When agents operate with broad permissions, any manipulation of the upstream can be converted into real-world actions. This makes access control a central security concern. Command injection also takes a different form here. The agent generates commands based on its reasoning, so if that reasoning is compromised, the resulting actions may still appear valid despite being harmful.
External tool outputs introduce another risk. If these systems return corrupted or misleading data, the agent may accept it without verification and incorporate it into its decisions. It is also becoming increasingly reliant on third-part tools and plugins adds to this exposure. If these components are compromised, they can affect behaviour without directly attacking the core system, creating a supply-side risk.
At this stage, the agent effectively operates as an insider. It holds legitimate credentials and interacts with systems in expected ways, making misuse harder to identify.
Application and Integration: System-Level Exposure
Agentic systems rarely operate in isolation. They are embedded in larger environments, interacting with identity systems, business logic, and operational workflows.
Access control becomes a major vulnerability. Agents tend to operate across multiple systems with various permission models, creating irregularities that can be exploited. Risks also arise from identity and delegation. In case an agent is operating on behalf of a user, then any vulnerabilities in authentication or session management can allow attackers to assume that authority.
Workflow execution amplifies these risks. Agents can initiate multi-step processes such as transactions, updates, or approvals. Manipulating a single step can change the result of the entire workflow. As integrations increase, so do the number of interaction points, making cumulative risk harder to track.
At this layer, failures are not isolated. They propagate into business operations, making consequences harder to contain.
Output and Action: Where Failures Become Visible
The output layer is where failures become visible, though they rarely originate there.
Data leakage has been a key concern. Agents may disclose information they are allowed to access, especially when tasks boundaries are not clearly defined. Misinformation and unsafe outputs are also important, particularly when outputs directly influence actions or decisions.
Generated code and commands introduce execution risk. If outputs are used without validation, errors or manipulations can have system-level effects. The shift towards autonomous action increases this risk, as small upstream deviations can lead to significant consequences without human intervention. This layer reflects symptoms rather than root causes. Addressing it alone does not reduce the underlying risk.
Beyond Layers: The Missing Dimension
A layered view helps, but it does not capture the full picture. Agentic systems are defined by continuous interaction across layers.
The key missing dimension is the runtime loop. Inputs shape reasoning, reasoning drives action, and actions feed back into both reasoning and memory. These cycles create feedback loops, where small manipulations may escalate over time. This also reduces observability. With multiple interacting components, it becomes difficult to trace cause and effect or identify where failures originate.
Supply chain dependencies add another layer of risk. Models, datasets, APIs, and plugins each introduce their own points of failure. A compromise at any of these points can propagate across the system. The attack surface also includes governance. Weak supervision, unclear responsibility, or excessive autonomy increase overall risk. Human control is not external to the system; it is part of its security.
Conclusion: Structuring the Attack Surface
Agentic AI expands the attack surface beyond traditional systems. It is both recursive and stateful. Risk does not just accumulate across layers; it moves and changes as the system operates.
Any useful representation must go beyond a linear stack. It should capture feedback loops, persistent state, and cross-layer dependencies that characterise the way these systems actually behave. The system is not a pipeline but a cycle. That is where both its capability and its risk emerge.
Related Blogs
Risk Management
The ‘Information Security Profile’ prioritises and informs cybersecurity operations based on the company's risk administration procedures. It assists in choosing areas of focus for security operations that represent the desired results for producers by supporting periodic risk evaluations and validating company motivations. A thorough grasp of the business motivations and safety requirements unique to the Production system and its surroundings is necessary in order to manage cybersecurity threats. Because every organisation has different risks and uses ICS and IT in different ways, there will be variations in how the profile is implemented.
Companies are currently adopting industry principles and cybersecurity requirements, which the Manufacturing Information is intended to supplement, not replace. Manufacturers have the ability to identify crucial operations for key supply chains and can order expenditures in a way that will optimise their impact on each dollar. The Profile's primary objective is to lessen and manage dangers associated with cybersecurity more effectively. The Cybersecurity Framework and the Profile are not universally applicable methods for controlling security risks for essential infrastructure.
Producers will always face distinct risks due to their distinct dangers, weaknesses, and tolerances for danger. Consequently, the ways in which companies adopt security protocols will also change.
Key Cybersecurity Functions: Identify, Protect, Detect, Respond, and Recover
- Determine
Create the organisational knowledge necessary to control the potential hazards of cybersecurity to information, systems, resources, and competencies. The Identify Function's tasks are essential for using the Framework effectively. An organisation can concentrate its efforts in a way that aligns with its approach to risk mitigation and company needs by having a clear understanding of the business environment, the financial resources that assist with vital operations, and the associated cybersecurity threats. Among the outcome characteristics that fall under this function are risk evaluation, mitigation strategy, the administration of assets, leadership, and the business environment.
- Protect
Create and put into place the necessary measures to guarantee the provision of crucial infrastructure amenities. The Protect Function's operations enable the limitation or containment of the possible impact of a cybersecurity incident. Instances of results Access Management, Knowledge and Instruction, Data Safety and Security, Data Protection Processes and Instructions, Repair, and Defensive Systems are some of the classifications that fall under this role.
- Detect
Create and carry out the necessary actions to determine whether a cybersecurity event has occurred. The Detect Function's operations make it possible to find vulnerability occurrences in an efficient way. This function's result subcategories include things like abnormalities and incidents, constant security monitoring, and identification processes.
- React
Create and carry out the necessary plans to address a cybersecurity event that has been discovered. The Response Function's operations facilitate the capacity to mitigate the effects of a possible cybersecurity incident. Within this Scope, emergency planning, interactions, analysis, prevention, and enhancements are a few examples of result categories.
- Recover
Create and carry out the necessary actions to uphold resilience tactics and restore any services or competencies that were hampered by a cybersecurity incident. In order to lessen the effects of a vulnerability incident, the Recovery Function's efforts facilitate a prompt return to regular operations. The following are a few instances of outcome subcategories under this role: communications, enhancements, and recovery planning.
Conclusion
The Information Security Profile, when seen in the framework of risk mitigation, offers producers a tactical method to deal with the ever-changing cybersecurity danger scenario. The assessment directs safeguarding operations prioritisation by recognising specific business reasons and connecting with corporate goals. The Profile enhances the cybersecurity standards and established industry guidelines by taking into account the differences in vulnerabilities and organisational subtleties among producers. It highlights the significance of a customised strategy, acknowledging that every business has unique risks and weaknesses.
The fundamental tasks of the Framework, to Identify, Protect, Detect, Respond, and Recover, serve as a thorough roadmap, guaranteeing a proactive and flexible approach to cybersecurity. The Profile's ultimate goal is to increase the efficacy of risk mitigation techniques, understanding that cybersecurity is a constantly shifting and evolving subject for the manufacturing sector.
References
- https://csrc.nist.gov/news/2020/cybersecurity-framework-v1-1-manufacturing-profile
- https://nvlpubs.nist.gov/nistpubs/ir/2020/NIST.IR.8183r1.pdf
- https://mysecuritymarketplace.com/reports/cybersecurity-framework-version-1-1-manufacturing-profile/
Executive Summary
A video is being shared on social media in which Prime Minister Narendra Modi can be heard saying that “a complete lockdown will be imposed from midnight to save the country.” Research by the CyberPeace found the viral claim to be misleading. Our probe revealed that the video is from March 2020, when PM Modi had announced a nationwide lockdown to curb the spread of COVID-19.
Claim:
An Instagram user shared the viral video on March 25, 2026. The link and archive link of the post are given below.
Fact Check:
To verify the claim, we conducted a keyword search on Google. However, we did not find any credible media reports confirming that such a lockdown announcement had been made recently. We then extracted keyframes from the viral video and performed a reverse image search using Google Lens. During this process, we found the same video on a YouTube channel, where it had been uploaded on March 24, 2020.
The viral portion of the clip appears around the 40-second mark in the original video.
Conclusion:
Our research found that the viral video is not recent. It dates back to March 24, 2020, when PM Modi announced a nationwide lockdown during the COVID-19 pandemic. The clip is being shared with a misleading claim.
.webp)
Introduction:
The Federal Bureau of Investigation (FBI) focuses on threats and is an intelligence-driven agency with both law enforcement and intelligence responsibilities. The FBI has the power and duty to look into certain offences that are entrusted to it and to offer other law enforcement agencies cooperation services including fingerprint identification, lab tests, and training. In order to support its own investigations as well as those of its collaborators and to better comprehend and address the security dangers facing the United States, the FBI also gathers, disseminates, and analyzes intelligence.
The FBI’s Internet Crime Complaint Center (IC3) Functions combating cybercrime:
- Collection: Internet crime victims can report incidents and notify the relevant authorities of potential illicit Internet behavior using the IC3. Law enforcement frequently advises and directs victims to use www.ic3.gov to submit a complaint.
- Analysis: To find new dangers and trends, the IC3 examines and examines data that users submit via its website.
- Public Awareness: The website posts public service announcements, business alerts, and other publications outlining specific frauds. Helps to raise awareness and make people become aware of Internet crimes and how to stay protected.
- Referrals: The IC3 compiles relevant complaints to create referrals, which are sent to national, international, local, and state law enforcement agencies for possible investigation. If law enforcement conducts an investigation and finds evidence of a crime, the offender may face legal repercussions.
Alarming increase in cyber crime cases:
In the recently released 2022 Internet Crime Report by the FBI's Internet Crime Complaint Center (IC3), the statistics paint a concerning picture of cybercrime in the United States. FBI’s Internet Crime Complaint Center (IC3) received 39,416 cases of extortion in 2022. The number of cases in 2021 stood at 39,360.
FBI officials emphasize the growing scope and sophistication of cyber-enabled crimes, which come from around the world. They highlight the importance of reporting incidents to IC3 and stress the role of law enforcement and private-sector partnerships.
About Internet Crime Complaint Center IC3:
IC3 was established in May 2000 by the FBI to receive complaints related to internet crimes.
It has received over 7.3 million complaints since its inception, averaging around 651,800 complaints per year over the last five years. IC3's mission is to provide the public with a reliable reporting mechanism for suspected cyber-enabled criminal activity and to collaborate with law enforcement and industry partners.
The FBI encourages the public to regularly review consumer and industry alerts published by IC3. An victim of an internet crime are urged to submit a complaint to IC3, and can also file a complaint on behalf of another person. These statistics underscore the ever-evolving and expanding threat of cybercrime and the importance of vigilance and reporting to combat this growing challenge.
What is sextortion?
The use or threatened use of a sexual image or video of another person without that person’s consent, derived from online encounters or social media websites or applications, primarily to extort money from that person or asking for sexual favours and giving warning to distribute that picture or video to that person’s friends, acquaintances, spouse, partner, or co-workers or in public domain.
Sextortion is an online crime that can be understood as, when an bad actor coerces a young person into creating or sharing a sexual image or video of themselves and then uses it to get something from such young person, such as other sexual images, money, or even sexual favours. Reports highlights that more and more kids are being blackmailed in this way. Sextortion can also happen to adults. Sextortion can also take place by taking your pictures from social media account and converting those pictures into sexually explicit content by morphing such images or creating deepfake by miusing deepfake technologies.
Sextortion in the age of AI and advanced technologies:
AI and deep fake technology make sextortion even more dangerous and pernicious. A perpetrator can now produce a high-quality deep fake that convincingly shows a victim engaged in explicit acts — even if the person has not done any such thing.
Legal Measures available in cases of sextortion:
In India, cybersecurity is governed primarily by the Indian Penal Code (IPC) and the Information Technology Act, 2000 (IT Act). Addressing cyber crimes such as hacking, identity theft, and the publication of obscene material online, sextortion and other cyber crimes. The IT Act covers various aspects of electronic governance and e-commerce, with providing provisions for defining such offences and providing punishment for such offences.
Recently Digital Personal Data Protection Act, 2023 has been enacted by the Indian Government to protect the digital personal data of the Individuals. These laws collectively establish the legal framework for cybersecurity and cybercrime prevention in India. Victims are urged to report the crime to local law enforcement and its cybercrime divisions. Law enforcement will investigate sextortion cases reports and will undertake appropriate legal action.
How to stay protected from evolving cases of sextortion: Best Practices:
- Report the Crime to law enforcement agency and social media platform or Internet service provider.
- Enable Two-step verification as an extra layer of protection.
- Keep your laptop Webcams covered when not in use.
- Stay protected from malware and phishing Attacks.
- Protect your personal information on your social media account, and also monitor your social media accounts in order to identify any suspicious activity. You can also set and review privacy settings of your social media accounts.
Conclusion:
Sextortion cases has been increased in recent time. Knowing the risk, being aware of rules and regulations, and by following best practices will help in preventing such crime and help you to stay safe and also avoid the chance of being victimized. It is important to spreading awareness about such growing cyber crimes and empowering the people to report it and it is also significant to provide support to victims. Let’s all unite in order to fight against such cyber crimes and also to make life a safer place on the internet or digital space.
References:
- https://www.ic3.gov/Media/PDF/AnnualReport/2022_IC3ElderFraudReport.pdf
- https://octillolaw.com/insights/fbi-ic3-releases-2022-internet-crime-report/
- https://www.iafci.org/app_themes/docs/Federal%20Agency/2022_IC3Report.pdf
