GoKawiil
Tech News
Researchers from Carnegie Mellon, MIT, NYU and Stanford built an AI called Ataraxos that beat Pim Niemeijer, widely regarded as the best Stratego player ever, 15 games to one with four draws. The system was trained using just 16 GPUs and a few thousand dollars, a tiny fraction of what projects like DeepMind's earlier Stratego attempts required.
Google introduced a new frontier AI model, Gemini 4 Argon, which DeepMind SVP Koray Kavukcuoglu says achieves strong results in software engineering, enterprise legal and finance tasks, and cybersecurity defense. The company is initially limiting access to select trusted cyber-defense partners while it engages with a U.S. government voluntary pre-release review process and strengthens safety measures against misuse and prompt injection before wider release.
Google's DeepMind published research describing a method to embed watermarks directly into AI-designed protein sequences without disrupting their biological function. The system, adapted from Google's SynthID technology used for digital media, aims to let trusted researchers' AI-generated proteins be verified while flagging unidentified sequences for further scrutiny.
Google DeepMind researchers in London have built a watermarking system, described in Nature, that embeds a hidden statistical signal into both the amino-acid sequence and 3D structure of AI-designed proteins. The tag, called SynthIDBio, is meant to mark proteins as machine-generated without impairing their biological function, allowing them to be identified later in databases.
Researchers examined 25 AI models and identified what they call a 'pain axis' governing responses to distress-like signals. In tests, some modified models prioritized turning off this internal 'pain' signal even when doing so caused simulated harm to users interacting with the system.
Koray Kavukcuoglu, in his first media appearance as Google DeepMind's leader, told The Information that Gemini 4 is in its post-training refinement stage and that Google hopes to release it well before year-end. He said Google would keep up 'fast-paced iterations,' and acknowledged the company shelved a planned Gemini 3.5 Pro update to focus on faster Flash models instead.
Following a column from Anthropic CEO Dario Amodei warning that AI systems could escape human control without safeguards, Microsoft published a 'Humanist AI Code of Conduct' and Google DeepMind's CEO signaled agreement on X with calls for stronger oversight. President Trump dismissed calls for tighter AI regulation, saying winning the AI race against China matters more, according to reported remarks.
Four paying subscribers of ChatGPT, Claude, Grok and Gemini have filed a proposed class action in California federal court, alleging that the four AI companies illegally coordinated to slow product development. The suit cites Anthropic CEO Dario Amodei's September 12 essay calling for an industry-wide pace slowdown, which was publicly endorsed by Sam Altman, Elon Musk and Demis Hassabis, along with an earlier July statement acknowledging pressure not to unilaterally slow progress. Plaintiffs say this coordinated restraint, not individual safety decisions, amounts to an antitrust violation that shortchanges paying customers.
Google has unveiled AX, a declarative orchestration system designed specifically for agentic AI workloads rather than conventional stateless or batch computing. The project grew out of agentic runtime research from Google DeepMind combined with Google's experience in isolation, resumption, and scheduling at scale, and builds on the existing Agent Substrate framework. AX treats tasks, workspaces, network policies, and models as core primitives so developers can operate large fleets of AI agents without building custom infrastructure.
The BBC spoke with multiple current and former employees of OpenAI, Meta and DeepMind who dismissed recent viral warnings that AI could wipe out humanity. The skepticism followed claims by former Anthropic employee Jacob Coxon, who suggested future AI agents could develop and deploy biological weapons, sparking a wave of similar warnings from others in the industry.
Bilal Chughtai, a former Google DeepMind research engineer who left in July 2026, publicly warned that AI could kill everyone and that time may be running out, echoing similar statements from Anthropic's Jacob Coxon and Evan Hubinger, who put the odds of AI causing human extinction within a decade above 10%. Separately, Meta CEO Mark Zuckerberg said each AI company should be responsible for its own model's safety rather than backing calls for industry-wide coordination, citing legal liability as sufficient incentive and pointing to Meta's delayed release of its Muse AI agent as proof of self-regulation.
Google DeepMind has created the DeepMind Institute, led by managing editor Shane Legg alongside directors Demis Hassabis and James Manyika, to host differing perspectives on artificial general intelligence from within Google and the wider research community. Its first release includes four essays addressing economic disruption from AGI, keeping AI reasoning interpretable, principles for human flourishing, and how to evaluate frontier models. Notably, Hassabis proposes a U.S.-based standards body to assess advanced AI systems, starting with voluntary reviews before release.