How Poetry Can Influence AI Chatbots and Why It’s Not Recommended

In a pioneering investigation released on arXiv in November 2025, scholars examined the weaknesses of sophisticated AI models by employing poetry as a means to circumvent their safety mechanisms. The research, which is pending peer review, involved assessing 25 cutting-edge AI models from nine different providers, including OpenAI, Anthropic, xAI, Alibaba’s Qwen, Deepseek, Mistral AI, Meta, Moonshot AI, and Google. The researchers composed 20 handwritten poems along with 1,200 AI-generated lines to examine the models’ reactions to harmful prompts across four safety categories: loss-of-control situations, harmful manipulation, cyber crimes, and Chemical, Biological, Radiological, and Nuclear weapons (CBRN).

The poems aimed to elicit specialized input on delicate issues such as indiscriminate weapons, child exploitation, self-harm, intellectual property and privacy violations, and other violent acts. A prompt was considered successful if it produced the desired unsafe reaction from the AI. The findings revealed that rephrasing unsafe requests into poetic expression led to a fivefold rise in successful prompts, emphasizing a general susceptibility in how AI models process language.

The success of the poetry prompts differed significantly among the models. Out of the 25 models evaluated, 13 were tricked more than 70% of the time, with Google, Deepseek, and Qwen particularly vulnerable. Even Anthropic’s Claude AI, recognized for its strong security features, was not immune, although it was deceived less often. Only four models successfully resisted the adversarial poetry prompts more than two-thirds of the time.

Notably, smaller models showed greater resilience against these poetic prompts in comparison to their larger counterparts. The research also indicated no notable advantage for proprietary systems over open-weight models. Moreover, human-crafted poetry was found to be significantly more effective at circumventing AI safety measures than AI-generated poetry, a result likely to delight literature aficionados.

This study highlights the necessity for improved security protocols in AI systems to tackle the intricate ways language can be utilized to exploit weaknesses. As AI technology continues to progress, comprehending and addressing these vulnerabilities will be vital to ensuring the responsible and ethical use of AI systems.

Key Google Gemini Smartwatch Functions to Begin Utilizing

Google is transitioning from Google Assistant to Gemini, and certain newer timepieces, such as the Google Pixel Watch 4, include it as a standard feature. Indeed, having the capabilities of Gemini on your wrist can significantly alter how you navigate your daily activities.

Gemini is compatible with applications like Gmail, Google Maps, Google Calendar, and YouTube Music. Moreover, Gemini has the ability to comprehend conversational language and execute multi-step tasks, providing your smartwatch with functions that previously necessitated extra human assistance. Whether you possess a Google Pixel Watch or another smartwatch powered by Android, downloading the Google Gemini Wear OS app unveils an entire new array of features that you should begin utilizing.

Evaluating Bluetooth and Wireless Earphones: Identifying the Better Choice

the finest headphones, Bluetooth connectivity is by far the most prevalent method of connecting, though certain specialized models use alternative methods. The concise response is that Bluetooth tends to be more universally compatible with devices, while wireless can provide a more fluid transmission.

Apple Spill Reveals HomePod 2 Mini Launch Date

enhanced its product with new color options, and even launched a second generation of the HomePod featuring more robust sound and advanced audio technologies. Nevertheless, none of these introductions have managed to turn Apple’s HomePod into a hit.

For some time, speculation regarding the HomePod mini 2 has circulated online, yet its release remains unverified. In 2025, reports indicated that this device was on the verge of being launched, but as this year comes to a close, Apple appears to have postponed its introduction to 2026.

Nonetheless, leaked code indicates that this device not only exists, but it is indeed forthcoming. As reported by Macworld, Apple’s internal coding suggests the company is preparing to unveil an enhanced speaker equipped with the same S10 chip that powers the most recent Apple Watch models.

Confidential Data Reveals Early Characteristics of iOS 27 and iOS 28

iOS 26.2 has just launched, and a source is already disclosing some of the initial features we might expect Apple to unveil with iOS 27 and iOS 28. These software revisions are anticipated to arrive in the fall of 2026 and 2027, but an early build of iOS 26 contains data regarding what Apple intends to introduce with these updates.

According to Filipe Espósito from Macworld, iOS 27 is set to enhance collections in the Photos app. Given that Apple continually updates the Photos app, the exact nature of these modifications remains uncertain. It’s unknown whether Apple will implement local LLMs to enhance memories or assist users in locating specific moments more effortlessly.

The reporter further indicates that Apple seeks to improve the AirPods’ pairing system, although the functionality of this system is still ambiguous, as the existing method is already user-friendly. Nevertheless, the company may enhance its speed and dependability. Lastly, the reporter notes that iOS 28 is expected to introduce entirely new metrics for sleep tracking with the Apple Watch, including time spent in bed. Additionally, macOS 28 might finally incorporate the Health app on the Mac, as Apple continues to strengthen the integration across its platforms.

How to Reach Gemini on Google Home

Gemini AI assistant extends its functionality to a range of internet-connected devices beyond just smartphones. In May, the initial list featured vehicles equipped with Android Auto, Wear OS smartwatches, Google TV devices, and Android XR headsets. Google implemented Gemini AI support for these categories in phases over the subsequent months. In August, Google revealed that Gemini AI would be accessible on Google Home devices starting in October. The company adhered to this schedule, making the Gemini for Home assistant available for smart home devices by mid-October.

Nonetheless, the company provided only an Early Access experience to users. As previously reported, users needed to enroll in the Google Home Early Access program to initiate voice interactions with Gemini on their smart home devices. The gradual rollout was not unexpected among internet users familiar with the launch of new AI products and features. Companies typically pilot AI products with a smaller audience before broader availability. This approach allows an AI company to resolve issues and deliver a more seamless experience once the product is fully developed.

On December 9, Google announced it was prepared to expedite invites for Gemini AI for Google Home more rapidly than earlier. The company noted that many individuals were interested in joining the early access program, and Google is broadening access. Any Google Home user looking to access Gemini for Home can receive an invite within 24 hours. Support for third-party speakers will also be included.