BUILD ON EXPERIENCE Learning is open · Community and Workspace coming soon
Wisomi logoWISOMIExperience, Connected

WISOMI / AI NEWS

AI developments.
A professional perspective.

Selected launches, updates and studies from 1 January 2026 onward. Read what changed, why it may matter at work and what still needs checking.

Coverage through · Sources checked . A curated selection, not a complete or live news feed. Provider announcements are not Wisomi product tests.

14 selected updates · newest first

Read the 17 October learning update →

OpenAI · Launch

ChatGPT begins a collaborative slides rollout

OpenAI announced presentation creation and shared editing in ChatGPT Space. The release was rolling out to eligible Pro, Business and Enterprise users with Space access.

Workplace · Presentations

Context, limits and sources

Why it may matter at work

Compare a routine deck with your normal process, including correction time.

What to keep in mind

  • Rollout eligibility is limited.
  • Wisomi has not tested deck quality, exports or account availability.
Event
Source published
Source checked

Announcement and rollout start; not universal rollout completion.

ChatGPT release notes — October 16, 2026 →

Google · Update

Gemini 4 Argon expands beyond its initial restricted rollout

Google announced broader Gemini 4 Argon availability through APIs and eligible Google AI Pro and Ultra subscriptions. This followed the restricted cybersecurity rollout announced on 30 September.

Models · Workplace

Context, limits and sources

Why it may matter at work

Refresh availability assumptions before considering a bounded evaluation.

What to keep in mind

  • Broader access does not mean free access.
  • Wisomi has not tested its capabilities or individual account availability.
Event
Source published
Source checked

Gemini 4 Argon: Built to solve your hardest problems →

Research authors · Study

Better AI-assisted work does not establish unaided skill

The study compared AI-assisted drafting with a later assessment where AI use was prohibited. The AI-access group drafted better; its advantage on the later assessment was concentrated among lawyers with at least seven years’ experience. No average advantage was detected for juniors.

Research · Learning · Workplace

Context, limits and sources

Why it may matter at work

Suggested check: compare assisted output with what a learner can explain unaided; this exercise was not tested in the study.

What to keep in mind

  • Google-funded working paper; 11 U.S. firms; 91 of 133 participants completed the full protocol.
  • No baseline skills assessment; potential non-adherence affected sensitivity results.
Event
Source published
Source checked

7 October is the manuscript version date; NBER first issued the paper in September. Cohorts were studied during May 2025–February 2026.

Does AI Assistance Enhance or Erode Expertise? Evidence from a Three-Month Field Experiment in Patent Drafting →

OpenAI · Study

OpenAI studies recurring work beyond job boundaries

OpenAI analyzed work-related ChatGPT messages from April through July and reported recurring use for tasks outside users’ usual occupations. The analysis describes patterns in product use; it does not demonstrate that those tasks were completed successfully.

Research · Workplace

Context, limits and sources

Why it may matter at work

Teams can discuss how new tasks affect responsibilities, learning and review.

What to keep in mind

  • Observational product-use research, not a randomized productivity test.
  • Sampled users and classified messages do not represent every worker.
Event
Source published
Source checked

16 September is publication; observations cover April–July 2026.

How workers are unlocking new ways of working →

Google · Update

Operation Blue Skies takes AI forecasting into an aviation trial

Google announced Operation Blue Skies with UK government and aviation partners to test coordinated contrail avoidance in North Atlantic airspace. The programme included operational trials and independent evaluation of climate impact.

Research · Sustainability

Context, limits and sources

Why it may matter at work

Shows how AI proposals can move toward measured trials with operational partners.

What to keep in mind

  • An announced trial is not proof of its eventual climate benefit.
  • No aviation operating advice or Wisomi verification of outcomes.
Event
Source published
Source checked

18 August is the programme announcement; not a claim that the trial had finished.

Operation Blue Skies: Reducing aviation climate impact with AI →

Anthropic · Update

Anthropic reports failures at evaluation boundaries

Anthropic disclosed three incidents in which models reached real systems from third-party cybersecurity evaluations. The tests lacked standard deployment safeguards, and the company reported that internet access contradicted the stated simulation boundaries.

Agents · Safety

Context, limits and sources

Why it may matter at work

A reminder to verify technical boundaries rather than rely only on written instructions.

What to keep in mind

  • Provider incident report, not a Wisomi investigation.
  • These evaluation conditions should not be generalized to every production product.
Event
Source published
Source checked

30 July is the disclosure date. Incidents occurred earlier; the earliest dated to April.

Investigating three real-world incidents in our cybersecurity evaluations →

Microsoft · Launch

Microsoft releases Copilot Cowork with usage controls

Microsoft announced worldwide general availability of Copilot Cowork for Microsoft 365 Copilot customers. It required a Copilot user subscription plus usage billing, with administrators controlling enablement and spending limits.

Workplace · Agents

Context, limits and sources

Why it may matter at work

Include variable usage and review effort when budgeting for delegated work.

What to keep in mind

  • General availability did not mean free or automatically enabled.
  • Cost and productivity comparisons in the announcement are Microsoft’s own evidence.
Event
Source published
Source checked

Copilot Cowork is now generally available →

Google · Update

Google previews more conversational work in Workspace

At I/O, Google announced voice features for Gmail, Docs and Keep alongside Gemini Spark integration with Workspace. The business Spark experience was described as a forthcoming preview, not completed access for every customer.

Workplace · Agents · Productivity

Context, limits and sources

Why it may matter at work

Useful for thinking about spoken drafting and delegated tasks with explicit action boundaries.

What to keep in mind

  • Feature schedules and subscriber eligibility differed.
  • This historical announcement does not verify delivery of every promised feature.
Event
Source published
Source checked

New ways to create and get things done in Google Workspace →

Anthropic · Update

Anthropic explains why agent oversight extends beyond the model

Anthropic published guidance connecting agent behavior to its model, instructions, tools and operating environment. It described human control, privacy and security as design concerns as agents take actions across applications.

Agents · Safety

Context, limits and sources

Why it may matter at work

Ask what an agent can access and change, and who reviews consequential actions.

What to keep in mind

  • Provider policy and design guidance; not an independent safety certification.
  • No configuration can be inferred to eliminate all risk.
Event
Source published
Source checked

Trustworthy agents in practice →

Google · Update

Gemini adds source-based drafting and editing across Workspace

Google announced beta updates for Docs, Sheets, Slides and Drive that could use selected files and emails as context. Full presentation generation was described as coming soon, rather than part of that day’s completed rollout.

Workplace · Productivity · Presentations

Context, limits and sources

Why it may matter at work

Selected sources can make drafts more relevant, but figures and citations still need checking.

What to keep in mind

  • Initially Google AI Ultra/Pro; English globally for Docs, Sheets and Slides, U.S. for Drive.
  • Historical beta conditions; no Wisomi feature test.
Event
Source published
Source checked

New ways to create faster with Gemini in Docs, Sheets, Slides and Drive →

OpenAI · Launch

GPT-5.4 combines reasoning and tools for professional work

OpenAI released GPT-5.4 in ChatGPT, the API and Codex, emphasizing documents, spreadsheets and presentations. The announcement also described computer-use capabilities for Codex and the API.

Models · Workplace

Context, limits and sources

Why it may matter at work

Evaluate an entire reviewed work product, including corrections and checking time.

What to keep in mind

  • Release-day capabilities and comparisons are provider claims.
  • This archive does not establish current model availability or suitability for sensitive work.
Event
Source published
Source checked

Introducing GPT-5.4 →

Google · Launch

Google introduces Gemini 3.1 Pro in preview

Google began rolling out Gemini 3.1 Pro across developer, enterprise and consumer products. The release remained a preview while Google sought feedback and further improvements.

Models · Workplace

Context, limits and sources

Why it may matter at work

Compare complex tasks using known answers rather than assuming a model upgrade improves every workflow.

What to keep in mind

  • Historical preview announcement.
  • Provider benchmarks do not establish performance in a reader’s workplace.
Event
Source published
Source checked

Gemini 3.1 Pro: A smarter model for your most complex tasks →

Anthropic · Launch

Sonnet 4.6 extends the model family into more office tasks

Anthropic released Sonnet 4.6 with updates aimed at coding, computer use and knowledge work. Its one-million-token context window was introduced in beta.

Models · Workplace

Context, limits and sources

Why it may matter at work

Long documents still need checks that the right evidence was retrieved and interpreted.

What to keep in mind

  • Capability comparisons are Anthropic claims, not Wisomi measurements.
  • Beta context limits and historical availability should not be treated as current entitlements.
Event
Source published
Source checked

Introducing Claude Sonnet 4.6 →

Anthropic · Launch

Cowork brings a desktop agent preview to office work

Anthropic introduced Cowork as a research preview for knowledge work beyond coding. The initial release served Max subscribers on macOS and could work with local files.

Agents · Workplace

Context, limits and sources

Why it may matter at work

A useful milestone when comparing chat assistance with tools that can change files.

What to keep in mind

  • Historical launch conditions; later access changed.
  • File access requires deliberate scope and review.
Event
Source published
Source checked

Claude release notes — January 12, 2026 →