EU AI Clock, Gemini Deep Think, and a Nobel Prize Hire
As EU AI Act transparency rules take effect in six weeks, Google launches a new reasoning mode and a Nobel laureate joins Anthropic.
By BINA Editorial
As the EU's artificial-intelligence transparency clock counts down to 2 August, this week's most significant developments span model reasoning, open-source security, scientific talent, and the future of work.
EU's Article 50 Clock: Six Weeks to Chatbot Disclosure Rules
On 2 August 2026 — 37 days from today — the EU AI Act's Article 50 transparency obligations become enforceable for the first time across all 27 member states. Chatbot providers must inform users they are talking to an AI; generative-AI providers must ensure synthetic content is detectable; platforms must clearly label deepfakes and AI-generated text published on matters of public interest. Companies that wish to benefit from a presumption of compliance should sign the European Commission's voluntary Code of Practice on AI-generated content by 22 July — less than four weeks away. Fines for non-compliance reach €15 million or 3% of global annual turnover, whichever is greater. High-risk AI obligations, deferred under the Digital Omnibus until December 2027, are a separate matter; the Article 50 transparency rules proceed entirely on schedule.
Google Makes Gemini 2.5 Pro's Deep Think Mode Public
Google released Gemini 2.5 Pro with Deep Think to the public on 22 June, calling it the most capable model it has ever shipped. Deep Think mode takes longer to respond than standard AI tools — it explores multiple reasoning paths and verifies its own logic before returning an answer. The model scored a bronze-equivalent result on the 2025 International Mathematical Olympiad problems and reached 82.4% on the GPQA Diamond science benchmark. Its context window holds 2 million tokens, double the capacity of any competing model currently available. Google positions the tool at the top of a three-tier stack: Gemini Flash for everyday use, Pro for professional tasks, and Deep Think for the hardest analytical problems.
OpenAI's Patch the Planet: Fixing Open-Source Security at Scale
OpenAI broadened its Daybreak cybersecurity programme on 22 June with two additions. First, the full release of GPT-5.5-Cyber — its most capable model for authorised security work, able to scan entire codebases, trace attack paths, and generate patches — became available to vetted defenders. Second, Patch the Planet, co-founded with security firm Trail of Bits and vulnerability platform HackerOne, deploys expert researchers equipped with Codex Security to assist maintainers of widely used open-source projects. More than 30 projects have committed to participate. The initiative marks a deliberate shift in emphasis: where earlier AI security tools mostly surfaced new vulnerabilities, Patch the Planet is designed to close them — addressing a persistent criticism that AI-powered discovery creates obligations that overwhelmed maintainers cannot meet.
Nobel Laureate John Jumper Leaves DeepMind for Anthropic
John Jumper, who shared the 2024 Nobel Prize in Chemistry for co-developing AlphaFold2, announced on 19 June that he is leaving Google DeepMind after nearly nine years to join Anthropic. AlphaFold2's ability to predict the three-dimensional structure of proteins from their amino acid sequences transformed drug discovery research and is regarded as one of the most consequential applications of machine learning in the life sciences. Anthropic has not disclosed the role Jumper will take, but the hire fits the company's stated interest in computational biology and medical applications of AI. His departure is the most prominent to leave DeepMind since the organisation became central to Google's AI strategy, and it signals the growing competition for frontier scientific talent.
PwC: AI Is Splitting the Labour Market in Two
PwC's 2026 Global AI Jobs Barometer, released this month and drawing on more than one billion job advertisements across 27 countries, found that AI is dividing the labour market into two distinct tracks. "Professionalised" roles — where AI handles routine tasks and employers place growing weight on human judgement, creativity, and leadership — are seeing twice the growth in job openings and 42% faster salary growth than "democratised" roles, where AI makes the work itself accessible to people without specialist training. The wage premium for workers with demonstrable AI skills rose to 62%, up from 57% a year earlier. A pattern the report calls "seniorisation" is visible at entry level: in AI-exposed sectors, junior postings demanding traditionally senior competencies have grown 35% since 2019, while other entry-level openings have contracted 10%.