> ## Content Index > Fetch the complete content index at: https://aitrading67.com/llms.txt > Use this file to discover other available public pages before exploring further. # AI & Work — Weekly Observatory · 27 June–3 July 2026 - URL: https://aitrading67.com/en/osservatorio-2026-07-03/ - Published: 2026-07-03T08:00:00.000Z - Updated: 2026-07-03T06:46:00.000Z - Description: US layoffs cooled in June, yet Challenger names AI as the top cited cause for a fourth straight month. The same week, Claude Sonnet 5 made agentic AI cheaper, California signed a landmark Anthropic deal, and Colorado's high-risk AI law took effect. - Author: Fabio Gentili - Tags: lang-en, observatory, AI & Work World Observatory · AI & Work AI & Work — Weekly Observatory · 27 June–3 July 2026 July 3, 2026 — Fabio Gentili observatoryAI & Work Editorial The week ending 3 July hands us two numbers that seem to contradict each other and in fact tell the same story. In its report published on 1 July, the outplacement firm Challenger, Gray & Christmas found that US employers announced 45,849 job cuts in June — down 53% from May and the lowest monthly total since December 2025\. Yet from the same report: for the fourth month running, artificial intelligence was the single most-cited reason for layoffs, named in 14,029 June announcements (31% of the month) and 101,743 positions year to date. Technology alone, with 139,156 cuts and 83% growth on 2025, accounts for nearly a third of all American layoffs this year. A cooling labour market in which AI's share of the cuts keeps rising is not a contradiction — it is the clearest sign yet that the restructuring is structural, not cyclical. And the week supplied the lever that makes it viable: on 30 June Anthropic launched Claude Sonnet 5, near-flagship agentic capability at an introductory $2/$10 per million tokens — what TechCrunch called "a cheaper way to run agents". The third movement was the tentative arrival of a policy response: RAISE US ($1B target, $500M committed), California's landmark Anthropic deal for 300,000 state workers, and Colorado's high-risk AI statute taking effect on 30 June. The week's real signal is their convergence. --- Weekly thematic blocks 🤖 01 · AI AI evolution — key developments High tension Claude Sonnet 5 ships near-Opus at Sonnet prices · Claude Science · Gemini 3.5 Pro slips again **Claude Sonnet 5 (30 June):** Anthropic's most agentic mid-tier model, default for all Free and Pro users from 1 July. Introductory pricing $2/$10 per M tokens through 31 August (then $3/$15). System Card: 63.2% agentic coding (vs Opus 4.8 69.2%), 80.4% Terminal-Bench 2.1 (+20.7 pts over Sonnet 4.6); first mid-tier Sonnet to edge the flagship on GDPval-AA v2. **Migration is not free:** extended thinking is always on by default, sampling parameters (temperature, top\_p) removed, and a new tokenizer yields 1.0–1.35× more tokens from the same text — three breaking changes that can silently break production agents. **Claude Science (30 June–1 July):** a dedicated research app for drug discovery, protein structure, genomics and computational biology, built on the \~$400M Coefficient Bio acquisition and the hire of Nobel laureate John Jumper. Puts Anthropic against OpenAI's science stack and DeepMind's Isomorphic Labs. **Google (30 June):** two image models shipped via AI Studio/Gemini API — Gemini 3.1 Flash Image ($0.50/$3.00) and Gemini 3 Pro Image — while Gemini 3.5 Pro missed its June GA (Polymarket 97% "no release") and slipped to July over token-consumption feedback. **OpenAI / xAI:** GPT-5.6 (Sol/Terra/Luna) previewed 26 June but restricted to \~20 government-vetted partners; Sol on Cerebras (up to 750 tokens/sec) planned for July, select customers. Grok 4.3 reached Amazon Bedrock ($1.25/$2.50); Grok 5 still training (Polymarket \~3% for 30 June). 💼 02 · WORK AI & work — displacement and transformation High tension Challenger: AI is the top cited layoff cause for a 4th month · 139,156 tech cuts YTD **Challenger June report (1 July):** 45,849 US cuts, −53% vs May's 97,006, −4% vs June 2025, lowest since Dec 2025\. H1 total 443,604 (−40% YoY, but 2025 was inflated by DOGE). AI is the leading stated reason for a fourth straight month. **AI's share:** AI cited in 14,029 June cuts (31% of the month) and 101,743 YTD (\~23% of all). Andy Challenger: "AI is the dominant force as companies are restructuring around it, automating roles, and reallocating budgets… The sector is being reshaped in real time." Firms named: Cloudflare, Snap, Block. **Tech epicentre:** 139,156 tech cuts through June, +83% on 2025 — nearly a third of the national total; June tech cuts 15,503 (down from 38,242 in May). **Worker expectations:** Anthropic's June Economic Index (\~9,700 users) finds >35% expect AI to do most/nearly all their tasks within 12 months; \~6 in 10 place next year's AI task-share higher than today's. The IMF flags a shock concentrated on entry-level roles. **Repricing knowledge work:** GitHub Copilot's first usage-based billing cycle closed 30 June with bills jumping 10×–50× ($29→$750; $50→$3,000). CPO Mario Rodriguez: "Copilot is not the same product it was a year ago." NYT (30 June): $180K SF engineers squeezed by a frontier-vs-middle barbell. 🎓 03 · SKILLS Automation and reskilling Medium tension RAISE US ($1B) and California's Anthropic deal embed reskilling into deployment **RAISE US (25 June):** Gina Raimondo and Eric Holcomb launch a bipartisan non-profit targeting $1B ($500M committed) to retrain US workers. Anchors: Amazon, Anthropic, Microsoft, OpenAI Foundation (also BofA, IBM, Cisco, Autodesk, GM, Eli Lilly). First states: Arkansas, Connecticut, Maryland, Utah. **Labour at the table:** AFL-CIO president Liz Shuler holds a board seat. Raimondo: "America has a technology strategy… It does not yet have a people strategy, and we cannot lead without one." **The paradox:** the retraining is largely funded by the firms building the technology that makes it necessary; a study of 23M federal-programme participants found reskilling rarely moved workers into less automation-exposed jobs. **California (29 June):** the largest US state AI deployment — Claude at 50% off for every agency and opting-in local government through a single procurement portal, with free workforce training and Anthropic engineering support. **Reskilling built-in:** the "Poppy" assistant, piloted with 2,800 staff across 67 departments, rolls out statewide in July; training is a contractual component of the deployment, not a separate programme after the cuts — 300,000 state workers as the test bed. 🚀 04 · ROLES New professions and opportunities Low tension A two-track market: AI-skill premium 56% · AI Engineer \~$206K · new public-sector and life-science roles **Two tracks (PwC 2026 Barometer):** "professionalised" roles (AI automates routine work, human judgement elevated) grow twice as fast as "democratised" ones, with salary growth 42% faster. **The AI-skill premium:** up to 56% (from 25% a year earlier); an AI-skilled candidate commands a \~23% higher advertised salary. AI Engineer remains the fastest-growing US title; MLOps base pay $90K–$257K; average AI Engineer \~$206K (+\~$50K YoY). **Where new roles appear:** California's deployment opens public-sector demand — government workflow designers, internal-training leads, automated code-security specialists. **Life-science pipeline:** Claude Science opens roles for computational biologists, experimental-data curators, and model-to-database integration engineers as three labs contest drug discovery. **The barbell caveat (NYT, 30 June):** the same technology minting trillion-dollar valuations compresses the technical middle. The new AI jobs pay well; the open question is whether they are numerous and accessible enough to offset the roles dissolving. ⚖️ 05 · ETHICS Ethics and ethical issues High tension Colorado's high-risk AI law live (30 June) · California's responsible-AI clause · EU toward 2 August **Colorado AI Act (in force 30 June):** first comprehensive US state statute on high-risk AI — mandatory risk-management programme, annual impact assessments, disclosure for "consequential" decisions, and a resident's right to appeal an AI-made decision (hiring, firing, credit, services). **California's responsible-AI framing:** the Newsom–Anthropic deal is the first major contract under the March 2026 executive order requiring vendors to show bias-prevention, civil-rights and misuse safeguards. CIO Chris Given said the federal DoD supply-chain risk designation "just didn't come up". **EU AI Act — 2 August:** Article 50 transparency for AI-generated content, GPAI penalty powers and market-surveillance authority all activate; Annex III high-risk obligations deferred to 2 December 2027. **The "enforcement gap":** many member states have yet to designate national competent authorities, and it is unresolved whether the Digital Omnibus will be adopted before 2 August. Publishers and the EBU warn draft deepfake guidance could over-label trusted media. ⚠️ 06 · RISKS AI risks High tension "Months, not years" cyber warning · Squidbleed & LiteLLM CVEs · IPO-bubble and monoculture risk **Cyber, "months not years":** the Five Eyes joint statement (22 June) frames the week — "The timeline is not years, it is months". CISA cut the mandatory federal patch deadline to three days, citing the AI threat directly. **Machines finding machine-era bugs:** "Squidbleed" (CVE-2026-47729), a 29-year-old Squid-proxy memory leak, was found by Claude Mythos 5 under Anthropic's Project Glasswing (23,019 confirmed vulns across 1,000 OSS projects in month one, 90.6% confirmation). LiteLLM's CVE-2026-42271 (RCE exposing API keys) added to CISA's KEV on 27 June. **Financial/systemic risk:** the triple IPO season (SpaceX public since 12 June, OpenAI targeting September, Anthropic October) runs on valuations markets struggle to justify (OpenAI projects $14B operating losses for 2026). Mid-June, Yann LeCun warned of a "big bubble explosion" absent cost cuts and higher prices. **Concentration / monoculture:** six senior researchers left Google DeepMind in five months, wiping $269B off Alphabet in the week of 18–24 June; Alphabet's $84.75B compute raise (Berkshire $10B anchor) captures a race pooling capital and talent in three labs — the premise of "algorithmic monoculture". 🎙️ 07 · VOICES Leading voices Medium tension Bengio's "Scientist AI" paper (2 July) · Altman & Raimondo on RAISE US · Amodei's launch · Hinton, LeCun, Hassabis **Yoshua Bengio (2 July):** his non-profit LawZero published a paper setting out a mathematical framework for a "Scientist AI" — an architecture that makes honest predictions without goals of its own. Non-agentic AI moves from manifesto to formal technical proposal, in the very week machines exposed decades-old flaws. **Sam Altman:** spoke mainly through RAISE US (OpenAI Foundation is a funder): "Helping people through the shifts that AI may bring to the economy is one of the most important things to start thinking through now." GPT-5.6 stays gated; IPO targeted for September. **Dario Amodei:** let the moves speak — Sonnet 5 and Claude Science in one week trace a lab integrating its stack vertically ahead of an October IPO. Long-run thesis remains augmentation over pure substitution. **Geoffrey Hinton (Fortune, 1 June, still central):** "probably more worried" than when he left Google, AI "has progressed even faster than I thought"; estimates perhaps "1%" of AI work goes to safety; frontier models are giving rise to "beings" smarter than us, "soon". **LeCun & Hassabis:** LeCun keeps his LLM-dead-end thesis (JEPA) and mid-June "big bubble explosion" warning. Hassabis reaffirms AGI "around 2030" (2029 possible) and "foothills of the singularity", arguing empathy-heavy jobs stay human — while DeepMind fights talent attrition and races to ship Gemini 3.5 Pro. --- Focus / In-depth analysis Focus Cheaper agents, confirmed AI layoffs, a nascent people strategy: how the week's pieces fit --- Conclusions What the week tells us, and what to watch next The week of 27 June–3 July 2026 tightened a loop rather than opening a new front. Challenger's June report confirmed that AI is the leading cited cause of US layoffs for a fourth consecutive month even as the overall pace cooled — a structural, not cyclical, reshaping led by a technology sector now carrying nearly a third of national cuts. On the same days, Claude Sonnet 5 made near-frontier agentic capability materially cheaper to run, and a first layer of policy response took shape: RAISE US, California's Anthropic deal with built-in training, and Colorado's high-risk AI statute. Watchpoints for the next seven days: (1) Gemini 3.5 Pro's July general availability; (2) any GPT-5.6 release beyond \~20 government partners; (3) EU AI Act readiness as 2 August nears (Article 50, GPAI penalties, the enforcement gap); (4) early enterprise adoption of Sonnet 5 as an Anthropic IPO signal; (5) LawZero's "Scientist AI" reception in the safety community; (6) the next Challenger monthly, to see whether AI's cited share of layoffs keeps rising. --- Sources & references [ 01 Challenger, Gray & Christmas — June layoffs cool to 45,849; AI leads for a 4th month 1 July 2026 https://www.challengergray.com/blog/challenger-report-june-layoffs-cool-to-45849-down-53-from-may-ai-leads-reasons-for-fourth-consecutive-month/ ](https://www.challengergray.com/blog/challenger-report-june-layoffs-cool-to-45849-down-53-from-may-ai-leads-reasons-for-fourth-consecutive-month/) [ 02 HR Dive — Tech is nearly a third of US layoffs in H1 2026, Challenger finds 2 July 2026 https://www.hrdive.com/news/tech-layoffs-surge-83percent-h1-2026-challenger-ai-disruption/824320/ ](https://www.hrdive.com/news/tech-layoffs-surge-83percent-h1-2026-challenger-ai-disruption/824320/) [ 03 CFO Dive — Tech accounts for nearly a third of US layoffs in H1: Challenger 2 July 2026 https://www.cfodive.com/news/tech-layoffs-surge-83percent-h1-2026-challenger-ai-disruption/824260/ ](https://www.cfodive.com/news/tech-layoffs-surge-83percent-h1-2026-challenger-ai-disruption/824260/) [ 04 Anthropic — Introducing Claude Sonnet 5 30 June 2026 · official launch https://www.anthropic.com/news/claude-sonnet-5 ](https://www.anthropic.com/news/claude-sonnet-5) [ 05 TechCrunch — Anthropic launches Claude Sonnet 5 as a cheaper way to run agents 30 June 2026 https://techcrunch.com/2026/06/30/anthropic-launches-claude-sonnet-5-as-a-cheaper-way-to-run-agents/ ](https://techcrunch.com/2026/06/30/anthropic-launches-claude-sonnet-5-as-a-cheaper-way-to-run-agents/) [ 06 VentureBeat — Claude Sonnet 5 at a steep discount as Anthropic races toward IPO 30 June 2026 https://venturebeat.com/technology/anthropic-launches-claude-sonnet-5-at-a-steep-discount-to-its-top-model-as-the-company-races-toward-a-blockbuster-ipo ](https://venturebeat.com/technology/anthropic-launches-claude-sonnet-5-at-a-steep-discount-to-its-top-model-as-the-company-races-toward-a-blockbuster-ipo) [ 07 Cybernews — Anthropic launches Claude Science, Sonnet 5 30 June 2026 https://cybernews.com/ai-news/anthropic-ai-claude-science-sonnet-5-launch/ ](https://cybernews.com/ai-news/anthropic-ai-claude-science-sonnet-5-launch/) [ 08 Build Fast with AI — AI News Today, 1 July 2026 1 July 2026 · daily aggregate https://www.buildfastwithai.com/blogs/ai-news-today-july-1-2026 ](https://www.buildfastwithai.com/blogs/ai-news-today-july-1-2026) [ 09 Build Fast with AI — AI News Today, 30 June 2026 30 June 2026 · daily aggregate https://www.buildfastwithai.com/blogs/ai-news-today-june-30-2026 ](https://www.buildfastwithai.com/blogs/ai-news-today-june-30-2026) [ 10 CryptoBriefing — Google delays Gemini 3.5 Pro to July 2026 June 2026 https://cryptobriefing.com/google-delays-gemini-35-pro-launch-to-july-2026/ ](https://cryptobriefing.com/google-delays-gemini-35-pro-launch-to-july-2026/) [ 11 CBS News Sacramento — California signs deal to bring Claude AI to state agencies 29 June 2026 https://www.cbsnews.com/sacramento/news/california-claude-deal-gavin-newsom/ ](https://www.cbsnews.com/sacramento/news/california-claude-deal-gavin-newsom/) [ 12 Rockefeller Foundation — RAISE US launches with employers and bipartisan governors 25 June 2026 https://www.rockefellerfoundation.org/news/raise-us-launches-uniting-nations-leading-employers-and-bipartisan-governors-behind-american-workers/ ](https://www.rockefellerfoundation.org/news/raise-us-launches-uniting-nations-leading-employers-and-bipartisan-governors-behind-american-workers/) [ 13 Axios — Anthropic joins Gina Raimondo's AI labor efforts 25 June 2026 https://www.axios.com/2026/06/25/anthropic-labor-market-ai-jobs-crisis ](https://www.axios.com/2026/06/25/anthropic-labor-market-ai-jobs-crisis) [ 14 GitHub Blog — GitHub Copilot moves to usage-based billing June 2026 https://github.blog/news-insights/company-news/github-copilot-is-moving-to-usage-based-billing/ ](https://github.blog/news-insights/company-news/github-copilot-is-moving-to-usage-based-billing/) [ 15 PwC — 2026 Global AI Jobs Barometer (two-track labour market) 15 June 2026 https://www.pwc.com/gx/en/news-room/press-releases/2026/pwc-2026-ai-jobs-barometer.html ](https://www.pwc.com/gx/en/news-room/press-releases/2026/pwc-2026-ai-jobs-barometer.html) [ 16 Cybersecurity Dive — Five Eyes warn AI-fueled threats require action now 22 June 2026 https://www.cybersecuritydive.com/news/ai-cyberattacks-five-eyes-frontier-models-warning/823526/ ](https://www.cybersecuritydive.com/news/ai-cyberattacks-five-eyes-frontier-models-warning/823526/) [ 17 LawZero — A formal safety case for "Scientist AI" 2 July 2026 https://lawzero.org/en ](https://lawzero.org/en) [ 18 Fortune — The 'godfather of AI' on beings smarter than us, soon (Hinton) 1 June 2026 https://fortune.com/2026/06/01/godfather-of-ai-geoffrey-hinton-beings-smarter-than-us/ ](https://fortune.com/2026/06/01/godfather-of-ai-geoffrey-hinton-beings-smarter-than-us/) [ 19 TechTimes — Google DeepMind's coding pivot lost six researchers 28 June 2026 https://www.techtimes.com/articles/319219/20260628/google-deepminds-coding-pivot-lost-six-researchers-meta-openai-anthropic.htm ](https://www.techtimes.com/articles/319219/20260628/google-deepminds-coding-pivot-lost-six-researchers-meta-openai-anthropic.htm) This document is a journalistic review built exclusively on public sources from the reference period (27 June – 3 July 2026), with a few immediate antecedents explicitly dated. **CONFIRMED/REPORTED/EXPECTED** classifications reflect each source's verification level: announcements not yet made or not verified by post-event sources are marked "expected/planned/reported". Not investment, legal or regulatory advice.