The evening briefing.
Today across AI and tech: Trump's administration eyes an OpenAI stake and military AI, while OpenAI launches a new Lockdown Mode. Plus, Google finds a novel way to power its data centers.
AI Governance. The Trump administration is reportedly in discussions to acquire an equity stake in OpenAI, while also issuing a memo to place advanced AI capabilities directly into military hands. This move prevents companies from altering military-used AI models without approval. Nvidia CEO Jensen Huang cautioned against quantifying AI's existential risks, suggesting science fiction influences unhelpful discussions. Former White House AI advisor Sriram Krishnan is also departing to establish a new institution focused on shaping Trump's AI policy.
AI Security & Agentic Systems. OpenAI has introduced Lockdown Mode to bolster protection for sensitive data against prompt injection attacks, though some vulnerabilities may persist. Meta confirmed thousands of Instagram accounts were compromised by abusing its AI chatbot. In AI development, a Universal Memory Protocol is emerging for agent memory, and Computex 2026 highlighted the "Agentic PC Era." The "agent war" is also intensifying, focusing on building "joyous products."
AI Applications & Infrastructure. Meta has launched its own AI-generated clickbait news feed within the standalone Meta AI app, featuring AI-created topics, images, and text. In infrastructure, Google is exploring innovative energy solutions for its data centers, partnering with Voltus to aggregate power from thousands of US households into a virtual power plant. On the battlefield, a lone Ukrainian robot reportedly held a key position against Russian forces for 45 days, showcasing the increasing reliance on autonomous systems in military operations.
Hardware & Software Development. BenQ has introduced its new MA270S, a glossy 27-inch 5K display designed specifically for Mac users, challenging the prevalence of matte finishes in the external monitor market. In software development, the Python JIT project has been requested to pause its development by the steering council. This decision impacts the ongoing efforts to integrate a Just-In-Time compiler into Python.
Trump's memo puts 'most advanced AI' into military hands
A new memo from the Trump administration directs the military to incorporate the world's most advanced AI. It also prevents companies from altering AI models used by the military without prior approval.
OpenAI unveils Lockdown Mode to protect sensitive data from prompt injection attacks
OpenAI has introduced Lockdown Mode to protect sensitive data from prompt injection attacks in ChatGPT. The goal is to reduce the likelihood of sensitive data being shared, even if some vulnerabilities may still exist.
Meta confirms thousands of Instagram accounts were hacked by abusing its AI chatbot
Meta has confirmed that thousands of Instagram accounts were compromised. The hacks occurred through the abuse of its AI chatbot.
The Trump administration might take an equity stake in OpenAI
The Trump administration is reportedly in talks about potentially acquiring an equity stake in OpenAI. President Trump stated discussions are underway for deals where the American people can benefit from AI's success.
Sriram Krishnan is leaving his role as White House AI advisor
Sriram Krishnan is reportedly stepping down from his position as a White House AI advisor. He is expected to establish a new institution to continue influencing Trump's AI policy.
Nvidia CEO Jensen Huang on quantifying AI existential risks: "it's not helpful"
Nvidia CEO Jensen Huang expressed that quantifying the existential risks of AI is not helpful. He believes discussions are influenced by science fiction, easing concerns about prominent warnings from scientists.
Google may have found a sneaky way to get all the electricity it needs for its data centers
Google is partnering with Voltus to aggregate household energy devices into a virtual power plant. This initiative aims to support the electricity demand growth of its data centers.
Universal Memory Protocol – a shared format for agent memory
A new Universal Memory Protocol is being proposed as a shared format for agent memory. This initiative aims to standardize how AI agents store and access information.
Computex 2026: Are We Heading for the Agentic PC Era Yet?
Discussions at Computex 2026 explored whether the industry is moving towards an "Agentic PC Era." This concept suggests a future where personal computers are highly autonomous and AI-driven.
"Whoever builds the most joyous product wins": The agent war begins
The "agent war" has begun, with a focus on building the most "joyous product" as highlighted at Snowflake Summit 2026. This signifies a shift towards user experience in the development of AI agents.
Meta made its own AI-generated clickbait news feed
Meta has introduced an AI-generated "For You" section in its standalone Meta AI app, populating a list of clickbait-style stories. The topics, images, and text are all created by AI.
Hands-on: BenQ’s new MA270S is a glossy 27-inch 5K display made for Mac
BenQ has launched the MA270S, a new 27-inch 5K monitor featuring a glossy finish, designed to complement Mac devices. This display aims to offer an alternative to the matte finishes prevalent in the external monitor market.
Python JIT project was asked to pause development
The Python JIT project has been requested to pause its development by the steering council. This decision affects the ongoing efforts to integrate a Just-In-Time compiler into the Python language.
A lone Ukrainian robot held a key position against hordes of Russian attackers for 45 days
A Ukrainian unmanned ground vehicle (UGV) reportedly defended a critical position against Russian forces for 45 days. This incident highlights the increasing reliance on robotic systems in modern warfare.