Meta employees are actively protesting new software that tracks their mouse movements and keystrokes, fearing the data will train AI systems designed to replace human workers. Protest flyers in US and UK offices label the program an 'Employee Data Extraction Factory,' while the company claims the tool helps AI learn real-world tasks to boost productivity. This tension arises as Meta prepares to lay off 8,000 employees, representing a 10% cut of its workforce.
OpenAI has developed a custom sandbox environment to run its Codex AI safely on Windows, addressing security limitations of standard tools. Simultaneously, the company launched Daybreak, an initiative using Codex Security to identify and fix software vulnerabilities. These efforts aim to balance safety with the ability to perform complex developer tasks and improve digital security.
In the medical field, a new study found an AI tool matches radiologists 98% of the time in detecting pulmonary embolisms after analyzing over 32,000 scans. While the AI performed well, doctors still identified 15% of cases the software missed, highlighting the continued need for human review in critical diagnoses.
Security leaders are discussing the first zero-day exploit created entirely by artificial intelligence, marking a dangerous new level of sophistication in cyber threats. In response, Microsoft uses AI to generate realistic fake attack logs, helping security teams test defenses without exposing sensitive information. Additionally, seven global cyber agencies released a guide on software bills of materials to track components in AI systems and improve supply chain safety.
South Korea is exploring a 'people's dividend' plan to share AI wealth with citizens through basic income or social credits to prevent poverty. Meanwhile, research indicates AI chatbots can help reduce loneliness but lack the depth of human therapists, warning against exaggerated claims about their effectiveness in mental health support.
Key Takeaways
['Meta workers are protesting new software that tracks mouse movements and keystrokes for AI training.', 'Meta plans to lay off 8,000 employees, which staff fear the monitoring software will help automate.', 'OpenAI created a custom sandbox environment to run Codex AI safely on Windows computers.', 'OpenAI launched Daybreak to use Codex Security for finding and fixing software vulnerabilities.', 'A new AI tool matches radiologists 98% of the time in detecting pulmonary embolisms across 32,000 scans.', 'Doctors still identified 15% of pulmonary embolism cases that the AI tool missed.', 'Security experts discuss the first zero-day exploit created entirely by artificial intelligence.', 'Microsoft uses AI to generate realistic fake attack logs to help improve threat detection.', 'Seven global cyber agencies released a guide on software bills of materials for AI supply chain safety.', "South Korea is considering a 'people's dividend' plan to share AI wealth with all citizens."]Meta Workers Protest AI Tracking Amid Layoff Fears
Meta employees are protesting new software that tracks their mouse movements and keystrokes. Staff fear this data will train AI systems designed to replace human workers. Protest flyers appeared in US offices with slogans calling the program an 'Employee Data Extraction Factory.' The company claims the tool helps AI learn real-world computer tasks to improve productivity. This tension comes as Meta prepares to lay off 8,000 employees soon.
Meta Staff Organize Against Surveillance Before Cuts
Meta workers have launched a petition to stop the company from monitoring their computer usage for AI training. Leaflets posted in UK and US offices warn that employees are helping build tools that might replace them. A group in the UK is also trying to unionize with the United Tech and Allied Workers. The company says the monitoring helps AI agents perform everyday tasks better. These actions happen just before Meta plans to cut 10% of its workforce.
Study Shows Mixed Results on AI Helping With Loneliness
New research suggests AI can help reduce loneliness but media headlines often exaggerate the findings. Dr. Lance B. Eliot explains that while AI chatbots are accessible, they lack the depth of human therapists. Some people use AI for mental health advice, but it carries risks of giving bad guidance. The study shows AI is a mixed bag for fighting loneliness depending on how it is used. Experts warn against false claims that humans always outperform AI in this area.
AI Matches Doctors in Finding Pulmonary Embolisms
A new study found an AI tool matches radiologists 98% of the time in detecting pulmonary embolisms. Researchers analyzed over 32,000 scans and found high agreement between the software and human doctors. However, doctors still caught 15% of cases that the AI missed. The combined approach of human review and AI analysis improved overall accuracy. This highlights the need for doctors to stay involved in critical medical diagnoses.
South Korea Proposes AI Dividend to Fight Poverty
South Korea is exploring a plan to share AI wealth with all citizens to prevent poverty. The government is considering a basic income guarantee or social credits for economic contributions. This idea, called a 'people's dividend,' aims to help everyone benefit from AI growth. Experts warn that without such measures, AI could widen the gap between rich and poor. The country wants to ensure social mobility while embracing this new technology.
OpenAI Builds Custom Sandbox for Secure Windows Use
OpenAI created a special sandbox environment to run its Codex AI safely on Windows computers. Standard Windows tools were not secure enough for the open-ended tasks Codex needs to perform. The new elevated sandbox allows the AI to access user files while keeping the system secure. This solution balances safety with the ability to run necessary developer tools. It overcomes the limitations of previous methods like AppContainer and Windows Sandbox.
OpenAI Launches Daybreak to Fix Software Bugs
OpenAI has introduced Daybreak, a new initiative to find and fix software vulnerabilities using AI. The program uses a security agent called Codex Security to analyze code and suggest fixes. This tool aims to stop cyber attacks by addressing weaknesses before hackers can exploit them. It competes with similar projects like Claude Mythos to improve overall digital safety. Early testing shows the AI can effectively identify potential threats in codebases.
Security Leaders Warn of First AI-Made Cyber Attack
Security experts are discussing the first zero-day exploit created entirely by artificial intelligence. This new type of attack uses AI to find weaknesses in secure systems that humans might miss. Leaders say this marks a dangerous new level of sophistication in cyber threats. They urge companies to be proactive and develop new defenses against AI-powered attacks. The discovery shows that AI can be a double-edged sword in cybersecurity.
Microsoft Uses AI to Generate Fake Attack Logs
Microsoft is using AI to create realistic fake security logs to help improve threat detection. Real attack logs are hard to get because malicious activity is rare and gathering them is risky. The new AI tool translates attacker tactics into structured data without exposing sensitive information. This helps security teams test defenses and build better rules faster. The method supports Microsoft Defender customers in staying ahead of cyber threats.
Global Agencies Release AI Supply Chain Safety Guide
Seven global cyber agencies released new guidance on software bills of materials for AI systems. The guide helps organizations track components and data used in AI to improve security. It defines seven key areas including models, datasets, and infrastructure details. The G7 Cybersecurity Working Group created this resource to tackle risks in the AI supply chain. Experts say this list must be used with other security tools to be fully effective.
Sources
- Amid looming layoffs, Meta workers accuse company of turning staff into AI training data
- Meta Employees Push Back Against AI Training Surveillance
- The Truth About Those Misleading Headlines Proclaiming That Humans Outdo AI When It Comes To Combatting Human Loneliness
- Ai Matches Radiologists in Detecting Pulmonary Embolism
- An AI ‘People’s Dividend’ Will Ensure Common Poverty
- OpenAI's Windows Sandbox Solution
- OpenAI launches Daybreak, its AI security initiative competing with Claude Mythos
- What Security Leaders Say About the First AI-Developed Zero-Day Exploit
- Accelerating detection engineering using AI-assisted synthetic attack logs generation
- Global Cyber Agencies Issue New SBOMs for AI Guidance to Tackle AI Supply Chain Risks
Comments
Please log in to post a comment.