Author: Anmol Khatiwada

The European commission is making strict rules for cleaner code of practice for AI developers. European commission appointed 13 experts all around the globe to draft a Code of Practice for General Purpose AI (GPAI) systems which includes popular models like chatGPT.

The group of professionals include PhD holders in AI, professor of civil, commercial and information law, researchers, doctors for medicine research and its application and experts in deep learning. The main goal of this act is to help AI developers obey the AI Act. It will focus on key areas like risk management, transparency, governance etc.

The team is stacked with talented individuals around the world who are trying to create a fair code of practice for developers. The commission will be taking several meeting with various stakeholders such as industry leaders, academics and community groups, by considering such different perspectives this will ensure a more comprehensive code. The drafting is said to be divided into 4 workstreams, each concentrating on a specific aspect of code. The experts aim to complete the Code by April 2025, including revision time. Once, completed this act will ensure ethical use of AI technology.

Summary

Russian hackers are sneaking into your devices and stealing your personal stuff using fake AI sites.

According to a recent report, Russian hackers are using fake AI sites as a trap to steal crypto wallet details. Cybersecurity firm suspects this to be the work of FIN7, the famous hacking group. The AI sites unknowingly download malware into your device.

Malware like RedLine and Lumma Stealer is designed to collect login credentials and other sensitive information including one’s crypto wallet details. By luring users into their fake AI site they install this malware into their PC, giving direct access to their personal information.

Zach Edwards, a senior analyst at Silent Push, explains that the majority of people targeted in this scam are tech-savvy individuals interested in AI tools. The hackers have cleverly disguised these sites to look legit by making effective use of Search Engine Optimization (SEO) and professional website design. Most of these websites promise deepfake nude images, which is directly connected to a Dropbox link which obviously doesn’t work and leaves the user with only malware on their devices. Some examples of such sites include “aiNude.ai”, “easynude.website” and “nude-ai. pro”.

FIN7 has a long history of hacking and and financial fraud, they are connected to several ransomware gangs, including DarkSide and BlackMatter. Although many of these sites have been taken down, experts warn anyone who interacted with them to consider whether their data is safe.

Summary

This past September hackers have managed to steal over $120 million from some big profiles on cryptocurrency platforms. Peckshield verified more than 20 incidents occurring during the month. The hackers are targeting both centralized and decentralized platforms.

The biggest damage by hackers in September was dealt to BingX, a Singapore-based exchange that lost about $44 million in a single incident. Penpie, a decentralized finance protocol also suffered a $27 million breach, making it 2nd most damaged entity by crypto hackers. Indodax, a leading cryptocurrency exchange in Indonesia, also suffered about $21million breach. These three represent a significant amount of losses the crypto community had to bear because of hacking.

Many other platforms, such as DeltaPrime, Truflation, Shezmu, Onyx, BananaGun, Bedrock, Cut, etc. were hacked, totaling up to $28 million which makes the overall hacked money well over $120 million. Despite the growing popularity of DeFi, it has become a major target for cybercriminals due to their insufficient security protocols. Most of the platforms hacked are from Asia, it seems that the hackers have found a way to exploit security gaps in the region’s rapidly growing crypto markets.

With September’s hacking, the total amount stolen in 2024 has reached a huge amount of $409 million. These incidents raise serious concerns about the security of the crypto market, casting doubt on its reliability and safety for investors.

Summary

Chromia just leveled itself up by teaming up with Elfa AI. Elfa AI is made for giving real-time instant market tips straight from places like X (Twitter) and Telegram. You can trade directly on the app and also customize alerts according to your vibes.

Researchers at HP discovered the first instances of AI-written malware scripts being sent out in emails. Malware developers are now using AI to generate codes to speed up the process and accelerate the number of attacks. The AI involvement lets less skilled individuals into hacking which only expands the number of hackers trying to steal your data.

In a September report Wolf security team from HP, detailed how they discovered a variation of asynchronous remote access trojan (AsyncRAT) on an email sent to a client. However, AsyncRAT is a human-developed software used to fix contaminated devices. These are exploited by hackers to steal user data. This new version contains an injection method which is supposedly developed using generative AI. Generative AI has a history of creating phishing or deceptive sites to lure users into buying their malware but this is the first time it has assisted in creating malware.

The team first discovered the email when it was sent to a subscriber of HP’s Sure Click threat containment software. It was written in French, suggesting it was likely a file targeting French speakers. This software allows users to control infected hosts remotely by providing them with a user interface that can perform tasks on the user’s computer, AsyncRAT can be used to steal a crypto user’s private key or seed words, potentially leading to the loss of funds.

“The activity shows how GenAI [generative AI] is accelerating attacks and lowering the bar for cybercriminals to infect endpoints,” the HP report stated. Cybersecurity details are still grappling with the effects of AI advancement on security. Many believe after the new EU draft is passed, these activities are less likely to occur.

A recent research in the Nature Scientific Journal revealed that AI chatbots are making more mistakes over time as newer models emerge. One of the key reasons for this may be because AI models are optimized to always provide believable answers, and the seemingly correct answers are always prioritized. It doesn’t recheck whether the news it’s giving is accurate or not. AI chatbots are just brilliantly coded search algorithm, which answers based on what is most written over the internet.

These AI hallucinations are self-reinforcing meaning older large language models train newer large language models. This method has only led to model collapse and the improvement is very minimum. Editor and writer Mathieu Roy cautioned users not to rely too heavily on these tools and to always check AI-generated search results for inconsistencies.

There have been numerous examples of Google’s finest artificial intelligence drawing blatantly inaccurate images. Some examples of this include portraying Nazis as black people, creating inaccurate images of well-known historical figures. Incidents like this are far too common and are getting worse with new updates. One of the fixes to this issue can be changing these AI hallucinations by forcing AI models to conduct thorough research and provide sources for every single answer given to a user. However, these have been done and the problems still exist. Just in, HyperWrite AI CEO Matt Shumer announced a new 70B model that uses “Reflection-Tuning”- meaning it will analyze its own mistakes and adjust responses over time.

With such stubborn and somewhat flawed search algorithms, AI chatbots provide inaccurate information, suggesting it cannot replace humans anytime soon.

Messari Copilot, the new AI assistant by Messari is designed to provide fast and accurate answers to cryptocurrency-related questions without having to search through endless data.

Messari claims that the AI can pull together information of over 30 terabytes from a variety of sources. It not only researches through current market and prices but also podcasts and reports posted on different social media. The major plus point for this AI assistant is its speed, it gives answers based on data that were updated as recently as 15 minutes. Each response it provided with the source that information was retrieved from, this allows users to verify whether the information they are getting is trustworthy or not.

Jimmy Skuros, Messari’s VP of Product, shared his excitement in the press release. He believes this project to be very fruitful. However, Messari copilot is only available to Messari Pro and Enterprise users at present.