AI News Roundup – April 24, 2026
Investors rotate to China’s chipmakers as DeepSeek intensifies AI competition
Following the disruptive release of the DeepSeek-V4 model, global investors are frantically rotating funds into Chinese semiconductor companies as demand for computing power skyrockets
South China Morning Post
Apr 24, 2026
OpenAI releases GPT-5.5, bringing company one step closer to an AI ‘super app’
OpenAI has released GPT-5.5, a faster model with superior reasoning and agentic skills that advances the company’s vision of an all-in-one AI super app
Tech Crunch
Apr 23, 2026
Anthropic’s Mythos breach was humiliating
Anthropic suffered an embarrassing security breach when unauthorized users simply guessed the online location of its highly guarded, unreleased Mythos model
The Verge
Apr 23, 2026
Teaching AI models to say “I’m not sure”
MIT researchers have developed a training method that cures AI of overconfidence by teaching models to calculate and express uncertainty, reducing high-stakes hallucinations
Massachusetts Institute of Technology
Apr 22, 2026
New report highlights agency advantages of using smaller, open-source AI models
A new report urges federal agencies to adopt smaller, transparent, open-source AI models over massive black-box systems to cut costs and boost security
FedScoop
Apr 22, 2026
Google unveils chips for AI training and inference in latest shot at Nvidia
Google directly challenged Nvidia by unveiling two specialized Tensor Processing Units dedicated exclusively to highly efficient AI training and inference workloads
CNBC
Apr 22, 2026
House lawmakers get a chilling demo of ‘jailbroken’ AI
Lawmakers received a stunning closed-door briefing where researchers easily jailbroke popular AI chatbots to generate step-by-step instructions for catastrophic physical attacks
Politico
Apr 22, 2026
Google and AWS split the AI agent stack between control and execution
AWS and Google have split the enterprise AI agent stack, with AWS prioritizing rapid deployment while Google champions strict, system-level governance to prevent rogue behavior
Venture Beat
Apr 22, 2026
AI swarms could hijack democracy without anyone noticing
Researchers warn that highly coordinated AI swarms are subtly infiltrating online communities to manipulate global elections and deeply erode trust in democratic systems
Science Daily
Apr 20, 2026
Artificial neurons successfully communicate with living brain cells
Northwestern University engineers have successfully printed flexible artificial neurons capable of transmitting lifelike electrical signals to communicate directly with living brain cells
Science Daily
Apr 18, 2026
AI Talks 