Techin' Geek

On Wednesday, a worldwide coalition of law enforcement agencies dismantled a botnet comprising tens of thousands of compromised home and small business routers.

The initiative focused on SocksEscort, which provided paid proxy services and was based on a botnet of compromised routers utilised to perpetrate a range of crimes, including unlawfully accessing victims’ banking and cryptocurrency accounts, as well as submitting fraudulent unemployment insurance applications, according to an announcement released by the Department of Justice (DOJ) on Thursday. The DOJ reported that the offenses enabled by SocksEscort incurred millions of dollars in losses for Americans.  

According to Europol’s statement regarding the operation, the SocksEscort botnet reportedly breached over 369,000 routers and Internet of Things devices across 163 nations, and the infected routers “have been disconnected from the service.” The law enforcement agency indicated that SocksEscort facilitated ransomware, distributed denial of service (DDoS) attacks, and the spread of child sexual abuse material (CSAM).

“Users of the illicit service purchased licenses to exploit these compromised devices, concealing their original IP addresses to partake in various criminal actions,” Europol stated. “Once infected with the malware, owners of the modems remained unaware that their IP addresses were being exploited for illegitimate purposes.”

The official website of SocksEscort was supplanted by a message announcing the seizure as part of the law enforcement operation. 

The botnet consisted of approximately 280,000 routers since last January and was driven by malware named AVRecon, as per cybersecurity firm Black Lotus Labs, which monitored SocksEscort and collaborated with authorities in the takedown effort.

“This botnet represented a considerable threat, as it was exclusively marketed to criminals,” the company remarked in its post regarding the takedown. “Significantly, more than half of its victims were situated in the United States or the United Kingdom, allowing perpetrators to execute highly targeted operations.”

In 2023, Black Lotus Labs referred to SocksEscort as “one of the largest botnets targeting small-office/home-office (SOHO) routers observed in recent years.” 

At that time, cybersecurity journalist Brian Krebs noted that SocksEscort emerged in 2009 as a Russian-language service offering access to thousands of compromised computers.

Last week, Grammarly unveiled a contentious feature that employs AI to mimic editorial feedback, creating the illusion that users are receiving critiques from figures like novelist Stephen King, the late scientist Carl Sagan, or tech journalist Kara Swisher. However, Grammarly did not obtain consent from the numerous experts featured in this function, referred to as “Expert Review,” to utilize their names.

Julia Angwin, one of the affected writers and a journalist, has initiated a class action lawsuit against Superhuman, the company that owns Grammarly, claiming that the organization infringed on the privacy and publicity rights of her and the other writers it represented. This class action allows other writers to join Angwin in her legal action.

“After decades of perfecting my craft as a writer and editor, I am disheartened to learn that a tech firm is profiting off a fraudulent version of my hard-earned knowledge,” Angwin expressed in a statement.

The irony of this situation is palpable — Angwin has dedicated her career to investigating how tech companies affect privacy. Other critics of such technology, including prominent AI ethicist Timnit Gebru, were also featured in Grammarly’s “Expert Review.”

The “Expert Review” feature, offered exclusively to subscribers at $144 annually, unfortunately falls short of its promise to provide meaningful feedback.

Casey Newton, the creator and editor of the tech newsletter Platformer and another individual impersonated by Grammarly, input one of his articles into the system and received feedback from Grammarly’s representation of tech journalist Kara Swisher. The feedback generated by Grammarly’s imitation of Swisher was so indistinct that it raises the question of why the company would bother using these writers’ personas at all.

This is what Grammarly’s representation of Kara Swisher suggested to him: “Could you briefly compare how daily AI users versus AI skeptics articulate risk, creating a through-line readers can follow?”

Newton conveyed the message from the AI’s rendition of Kara Swisher to the actual, living Kara Swisher.

“You greedy information and identity thieves better prepare for me to unleash my full McConaughey on you,” Swisher texted Newton (in reference to Grammarly). “Also, you suck.”

Grammarly has since taken down the “Expert Review” feature, as reported in a LinkedIn post by Superhuman CEO Shishir Mehrotra. While Mehrotra issued an apology, he continued to advocate for the concept behind the feature.

“Envision your professor refining your essay, your sales manager reworking a customer pitch, an insightful critic questioning your arguments, or a prominent expert enhancing your proposal,” he articulated. “For experts, this is an opportunity to forge that same pervasive connection with users, similar to what Grammarly has established.”

Tesla has officially become a licensed utility in the United Kingdom, as reported by The Wall Street Journal. The automotive and energy firm recently obtained a license from the Office of Gas and Electricity Markets, enabling it to provide electricity directly to residential and commercial users.

The company has been involved in electricity markets for some time. Its initial energy products, the Powerwall and Powerpack, debuted in 2015; however, it was only a year later, following its merger with SolarCity, that Tesla began to rapidly expand this division. In 2022, Tesla Electric was launched in Texas, permitting it to sell electricity directly to customers. Powerwall users can sell electricity from their batteries to engage in the company’s virtual power plant.

The new sector, branded Tesla Energy Ventures, will go head-to-head with current utilities in the U.K., such as EDF, E.ON, and Octopus Energy. The rivalry with Octopus is expected to be particularly captivating. Since its inception in 2015, Octopus has risen to be the largest utility in the country by prioritizing innovative software, renewable energy solutions, and imaginative marketing. Does that sound familiar? 

When Max Brodeur-Urbas co-established Gumloop in the middle of 2023, his aim was to assist non-technical staff in automating monotonous tasks using AI. At that point, the notion of AI agents was predominantly experimental and susceptible to inaccuracies.

As AI technology has advanced, Gumloop’s offerings have likewise evolved.

The company asserts that it now enables teams at organizations such as Shopify, Ramp, Gusto, Samsara, Instacart, and Opendoor to deploy dependable AI agents capable of autonomously managing intricate, multistep tasks, all without requiring any engineers.

Employees can share the agents they create with their teammates, fostering a compounding effect that speeds up internal automation. “They become hooked, they begin creating additional agents, and before long, the entire company is AI native,” Brodeur-Urbas shared with TechCrunch.

As businesses rush to embrace AI, Benchmark general partner Everett Randle believes that the secret to success lies in empowering every employee with AI capabilities, and Gumloop’s user-friendly agent builder exemplifies the type of tool that will unlock that potential.

That’s why Randle, who joined Benchmark last October from Kleiner Perkins, opted to spearhead a $50 million Series B funding for Gumloop. The agreement, which marks Randle’s first at his new firm, involved participation from Nexus VP, First Round Capital, Y Combinator, BoxGroup, The Cannon Project, and Shopify.

Although Gumloop was not actively looking for new funding, the startup determined that this was the year to “accelerate growth.” For Brodeur-Urbas, collaborating with Benchmark—the firm behind notable successes like eBay, Uber, and Dropbox—was an obvious choice.

While Brodeur-Urbas initially intended to “create a 10-person, billion-dollar company,” the increasing demand from corporate clients has driven him to establish a dedicated sales team and expand his engineering staff, he mentioned.

Gumloop is certainly not the sole contender striving to transform every knowledge worker into an AI agent-builder. The startup encounters tough competition from established automation platforms like Zapier and n8n, alongside specialized agent builders such as Dust. Even foundational AI labs are joining the competition. For example, Anthropic’s Claude Cowork enables users to develop autonomous agents without needing to write any code.

However, Randle is confident that Gumloop surpasses all its competitors. During his research, he found that at least one of the company’s customers had adopted Gumloop in a somewhat organic manner.

When Randle inquired with a CTO about how they selected Gumloop, the answer was revealing. The company had provided employees unrestricted access to Gumloop along with two other competitors. Six months later, the outcome was evident: Staff were engaging with Gumloop on a daily or weekly basis, while the competing platforms remained unused, Randle informed TechCrunch.

The reason Gumloop achieved such traction, according to Randle, is due to its low learning curve. “You can dive in and start creating agents and workflow automations right away,” he stated.

While numerous AI startups fear that foundational models will replicate their functionalities and make them redundant, Randle firmly believes that Gumloop’s model-agnostic strategy is exactly what will continue to draw in clients.

As models keep evolving, one may outperform another for specific tasks. Therefore, Gumloop offers the flexibility to select the model that is most appropriate for the job at any moment.

Another reason model independence is appealing, according to Randle, is cost. “Many enterprises possess OpenAI, Gemini, and Anthropic credits. They wish to utilize all of them,” he remarked.

His enthusiasm for the company ultimately hinges on the vast size of the opportunity.

“Enterprise automation is a tremendous pot of gold,” Randle stated. “I believe it’s the largest category in enterprise AI.”

Amazon’s AI assistant Alexa+ is receiving yet another new character. On Thursday, the firm revealed its intention to broaden its range of personality styles for users, introducing a “Sassy” option, designated exclusively for adults. Amazon notes that prior to utilizing the Sassy persona, users must complete additional security protocols within the Alexa app.

According to Amazon, this personality style will also be inaccessible when Amazon Kids is activated.

This new selection complements existing styles such as Brief, Chill, and Sweet, which were introduced last month.

When you enable the Sassy option in the Alexa mobile application, you are notified that the Sassy style incorporates explicit language, which necessitates a security verification. On iOS devices, this involved a Face ID scan.

The AI assistant articulated its persona as follows: “The Sassy style is constructed on one principle: help first, judge always. Every response is encapsulated with humor and a cleverly timed jab — it will address your inquiry; it will simply evoke a reaction first. Anticipate reality checks presented with charm, compliments that have an edge, and unexpected warmth. Equal-opportunity irreverence, no apologies. Genuine, incisive, and amusing — and somehow that proves to be more beneficial than just helpful.”

Alexa’s app also advised that this style might feature “mature subject matter.”

Nevertheless, further inquiry revealed that this is not Amazon’s equivalent to Grok’s adult AI companions. The AI assistant clarified that the new option will avoid topics such as explicit sexual content, hate speech, unlawful activities, personal attacks, or anything that may inflict harm on oneself or others.

This development exemplifies Amazon’s efforts to make Alexa+ more tailored, as it updates the assistant for the generative AI age. By presenting the assistant with various personalities — including one aimed at adults — Amazon is partaking in a broader trend in AI, where businesses are testing different tones, styles, and identities to make their assistants more engaging and personalized according to individual user preferences.

Amazon’s AI assistant Alexa+ is adopting a fresh personality. On Thursday, the company revealed it is enlarging its selection of personality styles for users, introducing a “Sassy” choice, designated for adults exclusively. Amazon notes that prior to utilizing the Sassy personality, users must complete additional security checks within the Alexa app.

According to Amazon, this personality option will also be unavailable when Amazon Kids is activated.

This new choice joins others like Brief, Chill, and Sweet, which were introduced last month.

When activating the Sassy option in the Alexa mobile app, you are alerted that the Sassy style incorporates explicit language, necessitating a security check. On iOS, this required a Face ID scan.

The AI assistant described its style as follows: “The Sassy style operates on one principle: help first, judge always. Each response is delivered with wit and a well-timed roast — it will address your question, just ensuring you feel something about it first. Anticipate reality checks served with charm, stinging compliments, and unexpected warmth. Equal-opportunity irreverence, no apologies. Honest, sharp, and humorous — and ironically, that’s more beneficial than merely being helpful.”

Additionally, Alexa’s app had cautioned that this style might encompass “mature subject matter.”

Nevertheless, further exploration revealed that this is not Amazon’s approach similar to Grok’s adult AI companions. The AI assistant clarified that the new option will refrain from delving into topics like explicit sexual content, hate speech, illegal actions, personal assaults, or anything that may endanger oneself or others.

This initiative exemplifies how Amazon is working to enhance the customization of Alexa+, as it modernizes the assistant for the generative AI landscape. By providing the assistant with varied personalities — including one tailored for adults — Amazon is tapping into a larger trend in AI, where firms are experimenting with tone, style, and characters to make their assistants more engaging and tailored to the individual preferences of users.