#FactCheck-Bangladesh Video Falsely Shared as Security Forces Action During West Bengal Elections 2026
Executive Summary
As West Bengal heads for vote counting on May 4, 2026, following the second phase of Assembly polling held on April 29, a video is being widely shared on social media. The clip shows security personnel baton-charging civilians, with users claiming it depicts force being used during the West Bengal Assembly Elections 2026. Research by CyberPeace Research Wing found that the viral claim is misleading. The video is actually from Bangladesh and is being falsely linked to the West Bengal elections to spread confusion.
Claim
A Facebook user named “Adv Mohd Salman” shared the clip on April 29, 2026, using Bengal-related hashtags and claiming that voters standing in line were beaten to influence the election outcome. The post alleged that free and fair voting rights were being suppressed.
Fact Check
To verify the claim, we closely examined the viral video. A vehicle visible in the footage had a registration number written in a non-Hindi script. Using Google Lens reverse image search, we found a matching image uploaded on Alamy on December 30, 2018. The image showed a military vehicle with the same script and registration style seen in the viral clip.
According to the description on the platform, the image was taken in Dhaka during Bangladesh’s national elections and showed Bangladeshi army personnel moving through a street near a polling station. This confirms that the viral footage is not related to the 2026 West Bengal Assembly elections.
Conclusion
Our research confirms that the video showing security personnel baton-charging civilians is from Bangladesh, not West Bengal. It is being falsely shared as footage from the 2026 West Bengal Assembly elections to mislead users.
Related Blogs
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.
Introduction
All citizens are using tech to their advantage, and so we see a lot of upskilling among the population leading to innovation in India. As we go deeper into cyberspace, we must maintain our cyber security efficiently and effectively. When bad actors use technology to their advantage, we often see data loss or financial loss of the victim, In this blog, we will shine light upon two new forms of cyber attacks, causing havoc upon the innocent. The “Daam” Malware and a new malicious app are the two new issues.
Daam Botnet
Since 2021, the DAAM Android botnet has been used to acquire unauthorised access to targeted devices. Cybercriminals use it to carry out different destructive actions. Using the DAAM Android botnet’s APK binding service, threat actors can combine malicious code with a legitimate application. Keylogging, ransomware, VOIP call records, runtime code execution, browser history collecting, incoming call recording, PII data theft, phishing URL opening, photo capture, clipboard data theft, WiFi and data status switching, and browser history gathering are just a few of the functions offered by the DAAM Android botnet. The DAAM botnet tracks user activity using the Accessibility Service and stores keystrokes it has recorded together with the name of the programme package in a database. It also contains a ransomware module that encrypts and decrypts data on the infected device using the AES method.
Additionally, the botnet uses the Accessibility service to monitor the VOIP call-making features of social media apps like WhatsApp, Skype, Telegram, and others. When a user engages with these elements, the virus begins audio recording.
The Malware
CERT-IN, the central nodal institution that reacts to computer security-related issues, claims that Daam connects with various Android APK files to access a phone. The files on the phone are encrypted using the AES encryption technique, and it is distributed through third-party websites.
It is claimed that the malware can damage call recordings and contacts, gain access to the camera, change passwords, take screenshots, steal SMS, download/upload files, and perform a variety of other things.
Safeguards and Guidelines by Cert-In
Cert-In has released the guideline for combating malware. These were issued in the public interest. The recommendations by Cert-In are as follows-
Only download from official app stores to limit the risk of potentially harmful apps.
Before downloading an app, always read the details and user reviews; likewise, always give permissions that are related to the program’s purpose.
Install Android updates solely from Android device vendors as they become available.
Avoid visiting untrustworthy websites or clicking on untrustworthy
Install and keep anti-virus and anti-spyware software up to date.
Be cautious if you see mobile numbers that appear to be something other than genuine/regular mobile numbers.
Conduct sufficient investigation Before clicking on a link supplied in a communication.
Only click on URLs that clearly display the website domain; avoid abbreviated URLs, particularly those employing bit.ly and tinyurl.
Use secure browsing technologies and filtering tools in antivirus, firewall, and filtering services.
Before providing sensitive information, look for authentic encryption certificates by looking for the green lock in your browser’s URL information, look for authentic encryption certificates by looking for the green lock in your browser’s URL bar.
Any ‘strange’ activity in a user’s bank account must be reported immediately to the appropriate bank.
New Malicious App
From the remote parts of Jharkhand, a new form of malicious application has been circulated among people on the pretext of a bank account closure. The bad actors have always used messaging platforms like Whatsapp and Telegram to circulate malicious links among unaware and uneducated people to dupe them of their hard-earned money.
They send an ordinary-looking message on Whatsapp or Telegram where they mention that the user has a bank account at ICICI bank and, due to irregularity with the credentials, their account is being deactivated. Further, they ask users to update their PAN card to reactivate their account by uploading the PAN card on an application. This app, in turn, is a malicious app that downloads all the user’s personal credentials and shares them with the bad actors via text message, allowing them to bypass banks’ two-factor authentication and drain the money from their accounts. The Jharkhand Police Cyber Cells have registered numerous FIRs pertaining to this type of cybercrime and are conducting full-scale investigations to apprehend the criminals.
Conclusion
Malware and phishing attacks have gained momentum in the previous years and have become a major contributor to the tally of cybercrimes in the country. DaaM malware is one of the examples brought into light due to the timely action by Cert-In, but still, a lot of such malware are deployed by bad actors, and we as netizens need to use our best practices to keep such criminals at bay. Phishing crimes are often substantiated by exploiting vulnerabilities and social engineering. Thus working towards a rise in awareness is the need of the hour to safeguard the population by and large.
Introduction
In April 2026, Anthropic revealed Claude Mythos, an artificial intelligence application capable of finding security flaws in computer networks more effectively than human beings. The corporation claimed to have found hundreds of thousands of substantially serious vulnerabilities in established desktop operating systems and web-based browsers that have not been used for at least 20 years. This news has greatly alarmed those responsible for leading financial organisations, banks, and governments throughout the world. Nevertheless, this news demonstrates a much larger problem: we do not have enough cybersecurity professionals trained to do this kind of work. At the current estimate, there are 4.8 million cyber security professionals short of what is needed globally. There is a need to develop different kinds of workforce training programs to help prepare these professionals as we continue to see the emergence of new AI technologies.
What Is Claude Mythos ?
Anthropic created Claude Mythos as part of its Claude AI system, competing against ChatGPT and Google Gemini. In April 2026, expert testing revealed Mythos excelled at identifying problems in legacy code and suggested exploitation methods. It found a vulnerability that had existed for 27 years. Because of these advanced capabilities, Anthropic restricted access through “Project Glasswing,” giving it only to 12 major tech companies and 40 organizations managing critical software. Canadian Finance Minister François-Philippe Champagne called it an “unknown unknown.” Andrew Bailey of the Bank of England said regulators needed to examine what Mythos could mean for financial attacks. The European Union raised concerns. India’s Finance Minister Nirmala Sitharaman warned at SEBI’s Foundation Day on April 25, 2026, that cybersecurity is the single most pressing challenge facing markets today. She stated a single successful cyberattack on a major exchange or large broker could disrupt markets nationally and shake public confidence for years. Sitharaman emphasized that AI tools make attacks faster, more adaptive, and autonomous, capable of discovering system vulnerabilities and manipulating code.
The Real Problem: Discovery Versus Fixing
Mythos highlights a fundamental mismatch in cybersecurity. Finding a vulnerability does not guarantee it will be fixed. Organizations face challenges patching systems. Many use obsolete technology, and updates can break dependent components. Organizations in developing nations often lack financial resources for repairs or downtime. Critical systems like hospitals, banks, and power grids cannot go offline. Before Mythos, human hackers found vulnerabilities slowly. Now AI tools find weaknesses faster than they can be fixed, creating a dangerous gap. Ciaran Martin, former head of the UK’s National Cyber Security Centre, explained that Mythos is “a really good hacker” against unprotected systems. Organizations following basic security practices—regular updates, strong passwords, network protection, trained staff can likely defend against it. The UK AI Safety Institute concluded Mythos poses the biggest threat to poorly defended systems, noting: “We cannot say for sure whether Mythos Preview would be able to attack well-defended systems.”
The Workforce Challenge
The Mythos announcement exposes the real problem: we lack enough trained cybersecurity workers. There is a global shortage of 4.8 million workers against a current workforce of 5.5 million. In AI security specifically, 34 percent of needed skills are missing. But the harder problem is that AI is changing needed skills. Entry-level jobs monitoring security alerts are being automated. These were traditional career starting points. Young people learned basic skills and moved to advanced roles. Now these positions disappear while new AI security jobs emerge for which nobody has training. Organizations cannot hire fast enough for new AI roles because few people have these skills. This leads to a vicious cycle. With fewer entry-level positions available, there will be fewer young adults entering the job market which results in even fewer workers with this skill set; thus, the shortage of qualified applicants increases; this thereby increases organizations’ vulnerability. Without action taken immediately, this issue will continue to worsen
Way Forward
- Clarify What Skills We Need
Governments and industry must work together to define what cybersecurity workers need in an AI world. Currently, aspiring professionals study networking, software, and vulnerability finding, but AI security training barely exists. Governments should work with universities and companies to clarify needed skills: understanding what AI tools can and cannot do in security, finding and fixing AI system problems.
- Support Workers Who Lose Jobs To Automation
Workers who find themselves losing their jobs due to automation will require government support. All too often without an alternative, these skilled and trained workers will leave their profession forever. The government will need to provide funding for training of displaced employees, support for those changing careers to become cyber security professionals.
- Create Clear Rules For AI Security Tools
When companies create powerful security tools, governments must understand their capabilities and risks. Companies should be required to thoroughly test tools before release, clearly explain what tools can do and their limitations, and explain safety and misuse prevention plans. Governments should monitor actual tool usage, not simply trust voluntary compliance.
- Focus On Basic Security First
Most attacks do not need advanced AI tools. They succeed because organizations have not implemented basic security. Some never update software, train employees, use strong passwords, protect data properly, or test defenses. Governments should require organizations, especially those managing critical systems, to implement these basics.
Conclusion
Claude Mythos matters not because it is a weapon of destruction, but because it forces hard questions: Do we have enough skilled workers? Are our systems well-protected? The answer is no. We face a shortage of 4.8 million cybersecurity workers and lack AI security training. Yet this is also an opportunity. Governments can invest in training, strengthen defenses, and create clear rules for AI security tools. Governments, organizations and educational institutions must collaborate to create viable Cybersecurity career pathways. We can act through either creating panic or creating a trained and prepared workforce to meet today’s challenges. The time is now.
References
- https://www.bbc.com/news/articles/crk1py1jgzko
- https://red.anthropic.com/2026/mythos-preview/
- https://www.anthropic.com/project/glasswing
- https://www.aisi.gov.uk/blog/our-evaluation-of-claude-mythos-previews-cyber-capabilities
- https://www.bsg.ox.ac.uk/people/ciaran-martin
- https://www.isc2.org/Insights/2024/10/Cybersecurity-Workforce-INSIGHTS-October-2024
- https://decrypt.co/364141/anthropic-claude-mythos-serious-threat-overhyped-ai-security-institute
- https://www.businesstoday.in/latest/economy/story/fm-nirmala-sitharaman-wants-sebi-regulated-entities-to-remain-exceptionally-vigilant-heres-why-527437-2026-04-25
- https://www.theweek.in/news/biz-tech/2026/04/25/sebi-38th-anniversary-cybersecurity-concerns.html
