#FactCheck-Viral Image of ‘New Iranian Banknote’ Featuring Khamenei Is Misleading; Likely AI-Generated

Introduction

The Senate bill introduced on 19 March 2024 in the United States would require online platforms to obtain consumer consent before using their data for Artificial Intelligence (AI) model training. If a company fails to obtain this consent, it would be considered a deceptive or unfair practice and result in enforcement action from the Federal Trade Commission (FTC) under the AI consumer opt-in, notification standards, and ethical norms for training (AI Consent) bill. The legislation aims to strengthen consumer protection and give Americans the power to determine how their data is used by online platforms. 

The proposed bill also seeks to create standards for disclosures, including requiring platforms to provide instructions to consumers on how they can affirm or rescind their consent. The option to grant or revoke consent should be made available at any time through an accessible and easily navigable mechanism, and the selection to withhold or reverse consent must be at least as prominent as the option to accept while taking the same number of steps or fewer as the option to accept.

The AI Consent bill directs the FTC to implement regulations to improve transparency by requiring companies to disclose when the data of individuals will be used to train AI and receive consumer opt-in to this use. 
The bill also commissions an FTC report on the technical feasibility of de-identifying data, given the rapid advancements in AI technologies, evaluating potential measures companies could take to effectively de-identify user data.

The definition of ‘Artificial Intelligence System’ under the proposed bill

ARTIFICIALINTELLIGENCE SYSTEM- The term artificial intelligence system“ means a machine-based system that—

1. Is capable of influencing the environment by producing an output, including predictions, recommendations or decisions, for a given set of objectives; and
2. 2. Uses machine or human-based data and inputs to

(i) Perceive real or virtual environments;

(ii) Abstract these perceptions into models through analysis in an automated manner (such as by using machine learning) or manually; and

(iii) Use model inference to formulate options for outcomes.

Importance of the proposed AI Consent Bill USA

1. Consumer Data Protection: The AI Consent bill primarily upholds the privacy rights of an individual. Consent is necessitated from the consumer before data is used for AI Training; the bill aims to empower individuals with unhinged autonomy over the use of personal information. The scope of the bill aligns with the greater objective of data protection laws globally, stressing the criticality of privacy rights and autonomy.

2. Prohibition Measures: The proposed bill intends to prohibit covered entities from exploiting the data of consumers for training purposes without their consent. This prohibition extends to the sale of data, transfer to third parties and usage. Such measures aim to prevent data misuse and exploitation of personal information. The bill aims to ensure companies are leveraged by consumer information for the development of AI without a transparent process of consent.

3. Transparent Consent Procedures: The bill calls for clear and conspicuous disclosures to be provided by the companies for the intended use of consumer data for AI training. The entities must provide a comprehensive explanation of data processing and its implications for consumers. The transparency fostered by the proposed bill allows consumers to make sound decisions about their data and its management, hence nurturing a sense of accountability and trust in data-driven practices.

4. Regulatory Compliance: The bill's guidelines call for strict requirements for procuring the consent of an individual. The entities must follow a prescribed mechanism for content solicitation, making the process streamlined and accessible for consumers. Moreover, the acquisition of content must be independent, i.e. without terms of service and other contractual obligations. These provisions underscore the importance of active and informed consent in data processing activities, reinforcing the principles of data protection and privacy.

5. Enforcement and Oversight: To enforce compliance with the provisions of the bill, robust mechanisms for oversight and enforcement are established. Violations of the prescribed regulations are treated as unfair or deceptive acts under its provisions. Empowering regulatory bodies like the FTC to ensure adherence to data privacy standards. By holding covered entities accountable for compliance, the bill fosters a culture of accountability and responsibility in data handling practices, thereby enhancing consumer trust and confidence in the digital ecosystem.

Importance of Data Anonymization

Data Anonymization is the process of concealing or removing personal or private information from the data set to safeguard the privacy of the individual associated with it. Anonymised data is a sort of information sanitisation in which data anonymisation techniques encrypt or delete personally identifying information from datasets to protect data privacy of the subject. This reduces the danger of unintentional exposure during information transfer across borders and allows for easier assessment and analytics after anonymisation. When personal information is compromised, the organisation suffers not just a security breach but also a breach of confidence from the client or consumer. Such assaults can result in a wide range of privacy infractions, including breach of contract, discrimination, and identity theft.

The AI consent bill asks the FTC to study data de-identification methods. Data anonymisation is critical to improving privacy protection since it reduces the danger of re-identification and unauthorised access to personal information. Regulatory bodies can increase privacy safeguards and reduce privacy risks connected with data processing operations by investigating and perhaps implementing anonymisation procedures.

The AI consent bill emphasises de-identification methods, as well as the DPDP Act 2023 in India, while not specifically talking about data de-identification, but it emphasises the data minimisation principles, which highlights the potential future focus on data anonymisation processes or techniques in India.

Conclusion

The proposed AI Consent bill in the US represents a significant step towards enhancing consumer privacy rights and data protection in the context of AI development. Through its stringent prohibitions, transparent consent procedures, regulatory compliance measures, and robust enforcement mechanisms, the bill strives to strike a balance between fostering innovation in AI technologies while safeguarding the privacy and autonomy of individuals.

References:

• https://fedscoop.com/consumer-data-consent-training-ai-models-senate-bill/#:~:text=%E2%80%9CThe%20AI%20CONSENT%20Act%20gives,Welch%20said%20in%20a%20statement
• https://www.dataguidance.com/news/usa-bill-ai-consent-act-introduced-house#:~:text=USA%3A%20Bill%20for%20the%20AI%20Consent%20Act%20introduced%20to%20House%20of%20Representatives,-ConsentPrivacy%20Law&text=On%20March%2019%2C%202024%2C%20US,the%20U.S.%20House%20of%20Representatives
• https://datenrecht.ch/en/usa-ai-consent-act-vorgeschlagen/
• https://www.lujan.senate.gov/newsroom/press-releases/lujan-welch-introduce-billto-require-online-platforms-receive-consumers-consent-before-using-their-personal-data-to-train-ai-models/

‍

Introduction

In the labyrinthine world of cybersecurity, a new spectre has emerged from the digital ether, casting a long shadow over the seemingly impregnable orchards of Apple's macOS. This phantom, known as SpectralBlur, is a backdoor so cunningly crafted that it remained shrouded in the obscurity of cyberspace, undetected by the vigilant eyes of antivirus software until its recent unmasking. The discovery of SpectralBlur is not just a tale of technological intrigue but a narrative that weaves together the threads of geopolitical manoeuvring, the relentless pursuit of digital supremacy, and the ever-evolving landscape of cyber warfare.

SpectralBlur, a term that conjures images of ghostly interference and elusive threats, is indeed a fitting moniker for this new macOS backdoor threat. Cybersecurity researchers have peeled back the layers of the digital onion to reveal a moderately capable backdoor that can upload and download files, execute shell commands, update its configuration, delete files, and enter states of hibernation or sleep, all at the behest of a remote command-and-control server. Greg Lesnewich, a security researcher whose name has become synonymous with the relentless pursuit of digital malefactors, has shed light on this new threat that overlaps with a known malware family attributed to the enigmatic North Korean threat actors.

SpectralBlur similar to Lazarus Group’s KANDYKORN

The malware shares its DNA with KANDYKORN, also known as SockRacket, an advanced implant that functions as a remote access trojan capable of taking control of a compromised host. It is a digital puppeteer, pulling the strings of infected systems with a malevolent grace. The KANDYKORN activity also intersects with another campaign orchestrated by the Lazarus sub-group known as BlueNoroff, or TA444, which culminates in the deployment of a backdoor referred to as RustBucket and a late-stage payload dubbed ObjCShellz.

Recently, the threat actor has been observed combining disparate pieces of these two infection chains, leveraging RustBucket droppers to deliver KANDYKORN. This latest finding is another sign that North Korean threat actors are increasingly setting their sights on macOS to infiltrate high-value targets, particularly those within the cryptocurrency and blockchain industries. 'TA444 keeps running fast and furious with these new macOS malware families,' Lesnewich remarked, painting a picture of a relentless adversary in the digital realm.

Patrick Wardle, a security researcher whose insights into the inner workings of SpectralBlur have further illuminated the threat landscape, noted that the Mach-O binary was uploaded to the VirusTotal malware scanning service in August 2023 from Colombia. The functional similarities between KANDYKORN and SpectralBlur have raised the possibility that they may have been built by different developers with the same requirements. What makes the malware stand out are its attempts to hinder analysis and evade detection while using grant to set up a pseudo-terminal and execute shell commands received from the C2 server.

The disclosure comes as 21 new malware families designed to target macOS systems, including ransomware, information stealers, remote access trojans, and nation-state-backed malware, were discovered in 2023, up from 13 identified in 2022. 'With the continued growth and popularity of macOS (especially in the enterprise!), 2024 will surely bring a bevvy of new macOS malware,' Wardle noted, his words a harbinger of the digital storms on the horizon.

Hackers are beefing up their efforts to go after the best MacBooks as security researchers have discovered a brand new macOS backdoor which appears to have ties to another recently identified Mac malware strain. As reported by Security Week, this new Mac malware has been dubbed SpectralBlur and although it was uploaded to VirusTotal back in August of last year, it remained undetected by the best antivirus software until it recently caught the attention of Proofpoint’s Greg Lesnewich.

Lesnewich explained that SpectralBlur has similar capabilities to other backdoors as it can upload and download files, delete files and hibernate or sleep when given commands from a hacker-controlled command-and-control (C2) server. What is surprising about this new Mac malware strain though is that it shares similarities to the KandyKorn macOS backdoor which was created by the infamous North Korean hacking group Lazarus.

Just like SpectralBlur, KandyKorn is designed to evade detection while providing the hackers behind it with the ability to monitor and control infected Macs. Although different, these two Mac malware strains appear to be built based on the same requirements. Once installed on a vulnerable Mac, SpectralBlur executes a function that allows it to decrypt and encrypt network traffic to help it avoid being detected. However, it can also erase files after opening them and then overwrite the data they contain with zeros..

How to keep your Apple computers safe from hackers

As with the best iPhones, keeping your Mac up to date is the easiest and most important way to keep it safe from hackers. Hackers often prey on users who haven’t updated their devices to the latest software as they can exploit unpatched vulnerabilities and security flaws.

Checking to see if you're running the latest macOS version is quite easy. Just click on the Apple Logo in the top right corner of your computer, head to System Preferences and then click on Software Update. If you need a bit more help, check out our guide on how to update a Mac for more detailed instructions with pictures.

Even though your Mac has its own built-in malware scanner from Apple called xProtect, you should consider using one of the best Mac antivirus software solutions for additional protection. Paid antivirus software is often updated more frequently and you often also get access to other extras to help keep you safe online like a password manager or a VPN.

Besides updating your Mac frequently and using antivirus software, you must be careful online. This means sticking to trusted online retailers, carefully checking the URLs of the websites you visit and avoiding opening links and attachments sent to you via email or social media from people you don’t know. Likewise, you should also learn how to spot a phishing scam to know which emails you want to delete right away.

Conclusion 

The thing about hackers and other cybercriminals is that they are constantly evolving their tactics and attack methods. This helps them avoid detection and allows them to devise brand-new ways to trick ordinary people. With the surge we saw in Mac malware last year, though, Apple will likely be working on beefing up xProtect and macOS to better defend against these new threats.

References 

• https://www.scmagazine.com/news/new-macos-malware-spectralblur-idd-as-north-korean-backdoor
• https://www.tomsguide.com/news/this-new-macos-backdoor-lets-hackers-take-over-your-mac-remotely-how-to-stay-safe
• https://thehackernews.com/2024/01/spectralblur-new-macos-backdoor-threat.html

As AI language models become more powerful, they are also becoming more prone to errors. One increasingly prominent issue is AI hallucinations, instances where models generate outputs that are factually incorrect, nonsensical, or entirely fabricated, yet present them with complete confidence. Recently, ChatGPT released two new models—o3 and o4-mini, which differ from earlier versions as they focus more on step-by-step reasoning rather than simple text prediction. With the growing reliance on chatbots and generative models for everything from news summaries to legal advice, this phenomenon poses a serious threat to public trust, information accuracy, and decision-making.

What Are AI Hallucinations?

AI hallucinations occur when a model invents facts, misattributes quotes, or cites nonexistent sources. This is not a bug but a side effect of how Large Language Models (LLMs) work, and it is only the probability that can be reduced, not their occurrence altogether. Trained on vast internet data, these models predict what word is likely to come next in a sequence. They have no true understanding of the world or facts, they simulate reasoning based on statistical patterns in text. What is alarming is that the newer and more advanced models are producing more hallucinations, not fewer. seemingly counterintuitive. This has been prevalent reasoning-based models, which generate answers step-by-step in a chain-of-thought style. While this can improve performance on complex tasks, it also opens more room for errors at each step, especially when no factual retrieval or grounding is involved.

As per reports shared on TechCrunch, it mentioned that when users asked AI models for short answers, hallucinations increased by up to 30%. And a study published in eWeek found that ChatGPT hallucinated in 40% of tests involving domain-specific queries, such as medical and legal questions. This was not, however, limited to this particular Large Language Model, but also similar ones like DeepSeek. Even more concerning are hallucinations in multimodal models like those used for deepfakes. Forbes reports that some of these models produce synthetic media that not only look real but are also capable of contributing to fabricated narratives, raising the stakes for the spread of misinformation during elections, crises, and other instances.

It is also notable that AI models are continually improving with each version, focusing on reducing hallucinations and enhancing accuracy. New features, such as providing source links and citations, are being implemented to increase transparency and reliability in responses.

The Misinformation Dilemma

The rise of AI-generated hallucinations exacerbates the already severe problem of online misinformation. Hallucinated content can quickly spread across social platforms, get scraped into training datasets, and re-emerge in new generations of models, creating a dangerous feedback loop. However, it helps that the developers are already aware of such instances and are actively charting out ways in which we can reduce the probability of this error. Some of them are:

1. Retrieval-Augmented Generation (RAG): Instead of relying purely on a model’s internal knowledge, RAG allows the model to “look up” information from external databases or trusted sources during the generation process. This can significantly reduce hallucination rates by anchoring responses in verifiable data.
2. Use of smaller, more specialised language models: Lightweight models fine-tuned on specific domains, such as medical records or legal texts. They tend to hallucinate less because their scope is limited and better curated.

Furthermore, transparency mechanisms such as source citation, model disclaimers, and user feedback loops can help mitigate the impact of hallucinations. For instance, when a model generates a response, linking back to its source allows users to verify the claims made.

Conclusion

AI hallucinations are an intrinsic part of how generative models function today, and such a side-effect would continue to occur until foundational changes are made in how models are trained and deployed. For the time being, developers, companies, and users must approach AI-generated content with caution. LLMs are, fundamentally, word predictors, brilliant but fallible. Recognising their limitations is the first step in navigating the misinformation dilemma they pose.

References

• https://www.eweek.com/news/ai-hallucinations-increase/
• https://www.resilience.org/stories/2025-05-11/better-ai-has-more-hallucinations/ 
• https://www.ekathimerini.com/nytimes/1269076/ai-is-getting-more-powerful-but-its-hallucinations-are-getting-worse/ 
• https://techcrunch.com/2025/05/08/asking-chatbots-for-short-answers-can-increase-hallucinations-study-finds/ 
• https://en.as.com/latest_news/is-chatgpt-having-robot-dreams-ai-is-hallucinating-and-producing-incorrect-information-and-experts-dont-know-why-n/
• https://www.newscientist.com/article/2479545-ai-hallucinations-are-getting-worse-and-theyre-here-to-stay/
• https://www.forbes.com/sites/conormurray/2025/05/06/why-ai-hallucinations-are-worse-than-ever/ 
• https://towardsdatascience.com/how-i-deal-with-hallucinations-at-an-ai-startup-9fc4121295cc/ 
• https://www.informationweek.com/machine-learning-ai/getting-a-handle-on-ai-hallucinations 

‍