A bimonthly American magazine on tech’s impact on culture, economy, and politics
OpenAI has said its unreleased model Astra may have reached a "Critical" capability for autonomous cyberattacks and has paused internal activities that do not meet heightened safeguards. The company has implemented isolated testing, universal monitoring for risky actions, and is working with government agencies and safety organisations to evaluate Astra's abilities.
Federal agencies and state officials have reported coordinated cyberattacks that have targeted internet-connected controllers at water and wastewater utilities in at least a dozen states since late July. Operators have had passwords changed and been locked out, forcing many plants to revert to manual control; officials say drinking water remains broadly safe while investigations continue.
The White House has authorised a program that will allow vetted U.S. private companies to carry out government-directed cyber surveillance and disruptive operations against overseas transnational criminal organisations. The memorandum has required Justice and Homeland Security oversight, a $1m escrow for participating firms, and a 60-day deadline for agencies to set participation rules.
Meta has rolled out updates to its AR glasses to deter covert filming, adding a stronger capture LED safeguard and hardware-based privacy measures. The moves come as regulators in the EU probe the glasses and venues consider bans; the company reports a small share of devices have been tampered with and vows ongoing updates.
Nvidia has invested in MediaTek’s AI chip ambitions, enabling NVLink Fusion connectivity and local AI computing. MediaTek’s shares have rallied as the partnership deepens MediaTek’s role in custom AI chips for data centers; the collaboration extends to PC and automotive platforms, signaling Nvidia’s continued influence in AI infrastructure.
Instagram has renamed its AI creator label to AI-generated profile and will label profiles with AI-generated people. The change aims to improve transparency as platforms face growing concerns over AI influencers. Accounts with AI-generated subjects may see reduced reach unless labeled; labels are not required for AI-assisted posts.
OpenAI and Anthropic are advancing their AI model releases ahead of potential IPOs, with Astra and Mythos/Fable updates designed to balance safety, performance, and enterprise data controls. OpenAI warns Astra safeguards may hinder legitimate work, while Anthropic is debuting higher-privacy safeguards and a zero-data-retention option for enterprise users.
A US-based watchdog has identified 332 paid ads containing child sexual abuse material (CSAM) on Meta’s Facebook and Instagram over the past 10 months, with a large share in August. The findings implicate ads promoting AI-generated CSAM and nudify apps, drawing responses from Indian authorities and Meta. Meta says it does not tolerate child exploitation and highlights its detection efforts.
Anthropic researcher Jacob Coxon has resigned and warned that Anthropic and OpenAI are "racing straight to self‑improving superintelligence" and "gambling with our lives," saying builders privately believe AI could kill humanity within the decade. Current and former employees, including Evan Hubinger, have publicly echoed similar existential concerns, prompting fresh calls in Washington for limits and oversight.
Massachusetts is tightening oversight of large data-center projects. Gov. Healey's executive order requires developers to source onsite clean power or fund nearby generation and to secure local community benefits before permitting. Applications for a data-center tax exemption are paused as regulators prepare to implement new rules.
Apple has introduced Reference Image, a feature for the iPhone 18 Pro that signs sensor data to produce an unalterable reference image. The tool lets users compare the original photo with later edits to verify authenticity, and will support Google SynthID in a software update later this year.
McClatchy has laid off over 90 reporters and staff across 17 publications, including The Sacramento Bee and The Miami Herald, affecting about 30% of unionized employees. Management says the cuts reflect a shift in investment away from subscriber-focused local news as print editions contract.
Iranian hard-liners have driven a July strike on three vessels in the Strait of Hormuz, derailing a U.S.-backed ceasefire. Mojtaba Khamenei’s absence fuels internal power struggles as the regime confronts renewed conflict and scrutiny over written messages attributed to the supreme leader.
Factory AI and Cognition remain at odds over confidential information and board-level conflicts. Vinod Khosla has publicly criticized Factory while investors debate the implications for valuations and competitive dynamics in AI coding tools.
Apple will introduce new controls to ensure that Full Disk Access for apps is granted only with explicit user action, citing rising risks as AI agents become more capable and autonomous. The move follows a wave of reports about Muse and similar concerns.
The Supreme Court is evaluating whether federal law preempts state and local lawsuits accusing oil companies of climate-related damages. Boulder, Colorado, leads a wave of cases seeking damages for disaster costs, with ExxonMobil and Suncor among the defendants. The decision could affect dozens of similar lawsuits nationwide.
OpenAI has coordinated new breakthroughs in mathematics, prompting the company to consult an independent panel of mathematicians to guide how results are released and engage with the mathematics community. The move follows public backlash over fast-paced, high-profile claims and aims to align future releases with community standards.
A Texas-based startup, Warhound, backed by investors including In-Q-Tel and Sequoia, has raised $200 million to develop genetically engineered military working dogs. Led by Palmer Luckey and Ben Lamm, the project aims to reduce Europe’s role as a supplier of combat dogs and create domestically bred, enhanced canines equipped with protective gear and sensor tech. The effort covers cloning, gene editing to improve health and performance, and integration with Anduril systems. The first batch is planned to enter service within two years.
UK ministers weigh incentives for vehicle-to-grid tech as part of the ZEV mandate review. The policy could reward carmakers for V2G adoption, while groups warn it may accelerate petrol-car sales if credits are misused. The move comes as grid balancing and consumer costs loom large in the transition to renewables.