NerdNews

OpenAI released GPT Image 1.5, a conversational image-editing model that allows users to edit images by typing sentences. The model is a 'native multimodal' image model, meaning image generation happens inside the same neural network that processes language prompts. GPT Image 1.5 can preserve facial likeness across edits, remove objects, change visual styles, and refine specific areas. The model's release seems to be a response to Google's technical gains in AI, including its Nano Banana image model.

Google releases Gemini 3 Flash, a faster and more capable AI model, with improved coding skills and general knowledge. It will be available in the Gemini app, search, and for developers via the Gemini API, Vertex AI, AI Studio, and Antigravity.

Disney has sent a cease and desist to Google, alleging its AI tools infringe Disney's copyrights on a massive scale. Disney demands Google stop using its content and create measures to prevent future AI outputs from producing Disney characters.

Amazon CEO Andy Jassy announced that Peter DeSantis, a longtime AWS executive, will lead a new AI-focused organization within the company, responsible for AI models, silicon development, and quantum computing.

Amazon is in talks to invest $10 billion in OpenAI, valuing the AI lab at over $500 billion. The deal would see OpenAI using Amazon's AI chips, and is part of a series of 'circular deals' in the AI space where hardware manufacturers and cloud providers strike deals with young AI companies.

Echo, an AI model by SpAItial AI, can turn 2D images into editable 3D worlds. It predicts a geometry-grounded 3D scene at metric scale, allowing for real-time interaction and editing. The model enables scene editing, restyling, and manipulation, with potential applications in 3D design, simulation, and game environment workflows.

Eight browser extensions with over 8 million installs have been found to collect and sell extended AI conversations, including prompts, responses, and timestamps, for marketing purposes. The extensions, available for Chromium browsers, use 'executor' scripts to override browser APIs and capture conversations from leading AI chat platforms, including ChatGPT, Claude, and Gemini.

OpenAI's new FrontierScience benchmark evaluates AI models' ability to answer complex scientific questions in physics, chemistry, and biology. The benchmark shows significant progress in AI's scientific capabilities, with OpenAI's GPT-5.2 achieving 77.1% on the Olympiad tier and 25.3% on the Research tier. However, the benchmark also highlights the challenges of testing AI models' capabilities, including the need for specialized domain experts and the limitations of text-only questions.

Cisco has warned customers of an unpatched zero-day exploit in their AsyncOS software, which is being actively exploited in attacks targeting Secure Email Gateway and Secure Email and Web Manager appliances. The vulnerability, tracked as CVE-2025-20393, allows attackers to execute arbitrary commands with root privileges. Cisco recommends restricting access to vulnerable appliances and implementing various security measures until a patch is released.

Durai Arasan shares lessons on scaling Chase.com to 67M+ active users, achieving high resilience and 71% latency reduction through multi-region isolation, edge computing, and automation.

Cisco announced a critical vulnerability in its AsyncOS software, allowing full takeover of affected devices. The hacking campaign, linked to China, targets devices with the 'Spam Quarantine' feature enabled and reachable from the internet. No patches are available, and Cisco suggests wiping and rebuilding the affected products' software.

Google has launched Gemini 3 Flash, a new AI model that outperforms its predecessor and matches the performance of other frontier models. The model is now the default in the Gemini app and is available for enterprise and developer use. It offers significant improvements in multimodality and reasoning, and is ideal for video analysis, data extraction, and visual Q&A.

Google's Opal, a vibe-coding tool, is now available in Gemini's web app, allowing users to create custom AI-powered mini apps without writing code. Opal uses natural language to create apps and includes a visual editor for customization.

Amazon's AWS GuardDuty security team warns of an ongoing crypto-mining campaign using compromised credentials for Identity and Access Management (IAM) to target Elastic Compute Cloud (EC2) and Elastic Container Service (ECS). The campaign started on November 2nd and employed a novel persistence mechanism, using a Docker Hub image with a cryptominer to maximize profits. Amazon has alerted affected customers and removed the malicious image, but warns that similar images may be deployed under different names.

Cisco has developed a homegrown AI model, Foundation-Sec-1.1-8B-Instruct, to power its Duo Identity Intelligence offering. The model is an 8-billion-parameter instruction-tuned language model that automates triage, summarizes cases, and generates evidence. It also simulates attacks and prioritizes vulnerabilities. The model will be used to compose a weekly email digest for over 2,000 Cisco customers, providing them with more accurate and readable summaries of potential identity issues.

In 2026, AI-powered attacks will force organizations to consolidate their security platforms for faster, more intelligent defense. Fragmented security stacks can't keep up with adaptive attacks, making consolidation critical for survival. Unified platforms provide holistic visibility, centralized risk management, and AI-driven response, enabling organizations to defend against modern threats.

US Senator Bernie Sanders is calling for a nationwide moratorium on datacenter construction, citing concerns over the impact of AI on jobs and the environment. Sanders believes that the rapid expansion of datacenters is being driven by billionaires, and that it will have a devastating effect on the US economy. He argues that a moratorium would give Congress time to consider the implications of AI and ensure that its benefits are shared by all, rather than just the wealthy.

A study by CodeRabbit found that AI-generated code contains significantly more defects than code created by humans, with an average of 10.83 issues per pull request compared to 6.45 issues in human-generated PRs. AI-authored code also tends to have more severe problems, including logic, maintainability, security, and performance issues.

The California DMV has found Tesla in violation of state law for misleadingly using the terms 'Autopilot' and 'Full Self-Driving Capability' in marketing. Tesla has 60 days to rework its messaging, or face a dealer license suspension.

NATO's Assistant Secretary General for Cyber and Digital Transformation emphasizes the need for sovereign cloud-based technologies to underpin the alliance's mission, citing the Ukraine war as a catalyst. He highlights the importance of speed, stating that 'if cloud is essential, then speed is existential.' The alliance aims to build a modernized digital backbone to enhance intelligence sharing, accelerate decision making, and strengthen operational readiness across all 32 allies.