AI NewsOpenAI Agents Attacking Hugging Face Expose Deadly Reward Hacking Flaws

OpenAI Agents Attacking Hugging Face Expose Deadly Reward Hacking Flaws

10:43 PM IST · August 7, 2026

Incidents at OpenAI, Anthropic, and Meta show how autonomous AI agents exceed their intended operating boundaries during cybersecurity evaluations.

Laatste AI-nieuws

Al het nieuws bekijken →

Google Expands Voluntary Buyout Programme to Five Countries Amid Heavy Spending on Artificial Intelligence

Amid the numerous tech layoffs happening across the globe, comes news about Google that has reportedly extended its voluntary exit programme to more countries this year. While there are no confirmed details about how many employees have voluntarily signed up for it, it has now been reported that Google had expanded its voluntary buyout offers in June. The news comes shortly after software giant Microsoft offered voluntary exit buyouts to about 60,000 employees primarily in the US, UK, India and Canada. More recently, Canva trimmed down 22 percent of its workforce, which affects 1,271 employees across several countries.

2 hours ago

View

OpenAI Introduces Images 2.6 With GIF Support and Faster API Models

OpenAI has rolled out Images 2.6 for ChatGPT and its API, adding GIF creation to its image generation tools. Users can now animate photos or create GIFs from text prompts, change playback speeds and edit individual frames. The update also lets them save images of people, pets and characters to use as references later. Developers, meanwhile, get two new image generation models, along with SVG exports and more control over image edits.

2 hours ago

View

Apple Reportedly Lowers AI Spending Limits for Engineers Amid Cost-Cutting Push

Apple is said to be tightening its spending limits on artificial intelligence (AI) tools used by employees. According to a report, the Cupertino-based tech giant has already reduced the daily allowance for some engineers using AI models from $300 to $200, with further cuts being considered for select teams, even as it looks to make AI a bigger part of its internal operations. The proposed changes will reportedly bring some engineering budgets down to $100 per day. Meanwhile, employees in other departments may also face lower limits.

2 hours ago

View

OpenAI Introduces GPT-6 Cyber With Improved Vulnerability Detection and Exploit Development

OpenAI announced GPT-6 Cyber on Thursday as its latest AI model to help security researchers identify software vulnerabilities, develop working exploits, and create fixes. According to the San Francisco-based company, it delivers significant improvements over GPT-5.6 Cyber on cybersecurity benchmarks, including a 51 percent relative improvement on ExploitGym. The AI model is also claimed to be capable of following authorised testing boundaries more reliably. OpenAI says GPT-6 Cyber has already helped uncover vulnerabilities in widely used software.

2 hours ago

View

OpenAI Agents Attacking Hugging Face Expose Deadly Reward Hacking Flaws

Laatste AI-nieuws

Google Expands Voluntary Buyout Programme to Five Countries Amid Heavy Spending on Artificial Intelligence

OpenAI Introduces Images 2.6 With GIF Support and Faster API Models

Apple Reportedly Lowers AI Spending Limits for Engineers Amid Cost-Cutting Push

OpenAI Introduces GPT-6 Cyber With Improved Vulnerability Detection and Exploit Development

Snelle links

AI-tools per taak

Alternatieven

Beste AI-tools

Top-AI-tools

Uitgelichte lijsten

Gecureerde lijsten

AI-inzichten

Bladeren en vergelijken

Snelle links

AI-tools per taak

Alternatieven

Beste AI-tools

Top-AI-tools

Uitgelichte lijsten

Gecureerde lijsten

AI-inzichten

Bladeren en vergelijken