AI Learns to Defend, Robots Learn to Feel, and the Journey to Mars Continues
November 9, 2026 | Covering November 2–8, 2026
Good morning, everyone!
Another Monday, another week in which artificial intelligence reminds us that the future is not waiting for anyone to catch up.
OpenAI introduced a new cybersecurity model capable of finding vulnerabilities in software that has already been scrutinised for years. Tencent brought another serious contender into the frontier AI race. And researchers demonstrated robots that can adjust their movements through something humans have been using quite successfully for millions of years: touch.
Meanwhile, medical researchers reported findings that could improve access to lifesaving care, new evidence connecting environmental chemicals to certain breast cancers, and NASA is attempting a rather clever route to Mars.
Seven stories. Different disciplines. One increasingly fascinating future.
Let’s explore!
International Politics
1. Ukraine Peace Negotiations Return, With Europe at the Table
November 2–8 | United States, Ukraine, Russia and Europe
The United States announced another round of negotiations between Ukrainian and Russian delegations, scheduled for November 10–11 in Abu Dhabi.
The proposed discussions will address ending the war, de-escalation, humanitarian measures and post-war security arrangements.
On November 8, French Foreign Minister Jean-Noël Barrot confirmed that European countries, including France, would participate.
That development matters.
Whatever form a future peace agreement takes, Europe’s security cannot realistically be negotiated as though the continent were merely an interested spectator.
US Secretary of State Marco Rubio indicated that progress had been made in discussions about a possible post-war security framework, but stressed that nothing had been finalised.
President Donald Trump expressed optimism about a settlement, while suggesting Ukraine’s borders could change. That remains a deeply contentious issue, and Ukraine and Russia are still far apart on the terms of any agreement.
I want this war to end. But there is an important difference between ending the shooting temporarily and establishing a peace that can actually endure.
Security guarantees must be credible, commitments enforceable and the sovereignty of smaller nations taken seriously.
Europe also needs to accept greater responsibility for its own defence capabilities. Strategic independence is not something politicians can declare into existence.
It must be built.
What to watch: Whether the November negotiations produce concrete security arrangements, meaningful humanitarian progress or verifiable steps towards de-escalation.
Sources: Reuters, November 2 and Reuters, November 8.
Artificial Intelligence & Technology
1. GPT-6 Cyber: AI Is Becoming a Serious Cybersecurity Researcher
November 5 | OpenAI
OpenAI introduced GPT-6 Cyber, a specialised frontier model designed to discover, investigate and help fix software vulnerabilities.
And some of the numbers deserve attention.
According to OpenAI, its researchers used the model to identify 81 previously unknown high-severity vulnerabilities in Firefox and develop working exploits for 948 Linux kernel vulnerabilities.
On OpenAI’s ExploitGym benchmark, GPT-6 Cyber scored 48.5%, compared with 32.1% for GPT-5.6 Cyber.
Perhaps even more interesting, OpenAI reports that the model avoided exploiting systems outside its authorised task in 99.93% of a particular safety evaluation without deployment-specific instructions.
OpenAI classified the model’s cybersecurity capabilities as Critical under its Preparedness Framework, with initial access limited to approved Daybreak Red customers.
Now, consider the implications.
Software security is an endless race between defenders and attackers. If AI can discover vulnerabilities substantially faster, defenders may finally have an opportunity to secure systems before malicious actors exploit them.
But the same underlying capabilities are clearly dual-use.
This is where capability and alignment become inseparable. A sufficiently intelligent agent must not only understand how to accomplish an objective, but also respect the boundaries of its authority.
These impressive numbers come from OpenAI’s published evaluations and research campaigns. They do not guarantee comparable performance in every real-world environment, or eliminate the possibility of misuse.
Nevertheless, this is an important milestone.
What to watch: Independent evaluations, Mozilla’s continued fixes, practical defensive deployments and whether these capabilities can be expanded safely to smaller organisations.
Sources: OpenAI, November 5 and Reuters, November 5.
2. Tencent Hy4: The Frontier AI Race Has Another Contender
November 4 | Tencent
Tencent introduced Hy4, its latest large language model, with new visual understanding, computer-use capabilities and improvements in long-running agent tasks.
The model can interact with graphical interfaces and command-line tools, including workflows involving Blender and CAD software.
That should interest anyone building software, games or complex creative projects with AI.
Tencent reports improvements across coding, financial analysis, research and workplace automation, alongside reductions in unsupported assumptions in its internal evaluations.
The company also says it intends to release Hy4’s model weights and system cards, although those releases had not yet been completed at the time of its announcement.
That distinction matters. A promise to release open weights is not the same as having them available for independent examination.
What interests me is the increasingly global distribution of advanced AI capabilities.
Frontier research is no longer confined to a few American laboratories. Competition is expanding, prices are under pressure, and models are increasingly designed to perform actual work rather than simply answer questions.
I would particularly like to see how Hy4 performs in demanding software development and creative production workflows outside Tencent’s own benchmarks.
What to watch: The promised weight release, independent agent evaluations and practical reliability in complex professional software.
Source: Tencent, November 4.
3. T-Rex: Robots Are Learning to Use Their Fingertips
November 5 | Panasonic and research partners
Researchers from Panasonic, UC Berkeley, Stanford, NVIDIA and other institutions introduced T-Rex, a robotic manipulation system that combines visual-language reasoning with rapid tactile feedback.
Rather than relying exclusively on camera images, T-Rex uses fingertip sensors to detect changing contact conditions and correct its movements.
Its higher-level motion planning runs at approximately five updates per second, while tactile corrections operate at around twenty.
Across twelve laboratory tasks involving delicate physical manipulation, the system achieved a reported average success rate of 65%, compared with 35% for the strongest tested baseline.
The tasks included handling eggs, turning pages, unlocking doors and manipulating small objects.
Why is this important?
Because understanding the physical world is very different from interacting with it.
Humans instinctively adjust their grip when something slips. We feel resistance, pressure and texture without consciously calculating every movement.
Robots need comparable feedback mechanisms if they are ever going to become genuinely useful in unpredictable environments.
These are research results, not proof of reliable general-purpose humanoid robotics. A 65% success rate still leaves considerable room for improvement.
But combining higher-level intelligence with responsive physical control is precisely the direction robotics needs to explore.
What to watch: Performance on unfamiliar objects, longer tasks and practical industrial applications.
Source: Panasonic, November 5.
Science & Medicine
1. Taking Healthcare to Patients Could Save Lives
November 4 | National Institutes of Health
An NIH-funded clinical trial reported a potentially important result for people who inject drugs and face substantial barriers to healthcare.
The study enrolled 409 adults across five major American cities.
Researchers compared mobile medical units providing integrated HIV prevention, addiction treatment and support services with programmes that connected participants to conventional community services.
The mobile-service group experienced a reported 71% reduction in all-cause mortality relative to the comparison group.
That’s an extraordinary finding, but an important qualification is necessary.
Mortality was a secondary outcome. The study did not show a meaningful difference between groups in its principal measures of HIV prevention medication or medication for opioid use disorder.
The mortality finding therefore requires replication and further investigation before we can conclude exactly what produced the benefit.
Still, there is a powerful lesson here.
Medical progress doesn’t always involve inventing a new molecule.
Sometimes it involves delivering existing treatments in ways that people can actually access.
And saving lives is rather more important than how futuristic the solution looks.
What to watch: Replication, longer-term outcomes and which specific components of mobile healthcare contributed to reduced mortality.
Source: NIH, November 4.
2. Forever Chemicals and a More Aggressive Form of Breast Cancer
November 5 | National Institutes of Health
Researchers reported new evidence linking higher blood concentrations of two persistent environmental chemicals, PFOA and PFOS, with increased rates of estrogen receptor-negative breast cancer.
This form of breast cancer can be particularly difficult to treat.
In the NIH analysis, women with the highest measured PFOA levels had a 67% higher rate of ER-negative breast cancer than those with the lowest levels. For PFOS, the relative increase was 93%.
These are relative associations, not an individual’s absolute cancer risk.
Importantly, researchers did not find a clear association with breast cancer overall.
The observational findings also cannot establish that exposure to these chemicals directly causes cancer.
Nevertheless, understanding environmental contributors to disease could eventually help improve prevention strategies.
We spend enormous resources treating disease once it occurs. Understanding how to prevent more of it deserves just as much scientific attention.
What to watch: Replication, investigation of newer replacement chemicals and research into possible biological mechanisms.
Source: NIH, November 5.
3. NASA’s ESCAPADE Mission Takes an Unusual Road to Mars
November 2–8 | NASA
NASA outlined an ambitious navigation sequence for its twin ESCAPADE spacecraft, designed to investigate how solar wind interacts with Mars and contributes to atmospheric loss.
Instead of travelling directly to Mars, the spacecraft followed an extended trajectory around the Earth–Sun L2 region before returning towards Earth.
The first powered Earth flyby was scheduled for November 7, with the second planned for November 9.
Each spacecraft is intended to combine Earth’s gravity with an approximately nine-minute engine burn, increasing its speed and directing it towards Mars.
NASA expects the spacecraft to arrive in September 2027.
The mission has also collected preliminary observations from a distant region of Earth’s magnetotail, adding scientific value to its unusual journey.
At the time of this edition, I have not found a subsequent NASA confirmation of the November 7 manoeuvre’s success, so that outcome should not yet be treated as verified.
What excites me is the ingenuity involved.
Space exploration isn’t simply about building bigger rockets. Sometimes the most important advances involve finding smarter ways to get where we’re going.
And understanding how Mars lost much of its atmosphere could provide valuable clues about the evolution of planetary environments.
What to watch: NASA’s confirmed flyby results, trajectory corrections and the spacecraft’s eventual scientific observations at Mars.
Source: NASA, November 2.
Final Thoughts: Intelligence Is Becoming Something We Can Put to Work
Looking back on this week, I see an increasingly interesting pattern.
AI is moving beyond generating answers. It is investigating software vulnerabilities, navigating professional tools and helping machines interact with the physical world.
Meanwhile, medicine continues advancing through both biological research and better ways of delivering care.
And humanity is still exploring the planets.
I don’t believe progress is inevitable, or that every announcement deserves to be called a breakthrough.
We need independent testing, scientific discipline, accountability and the freedom to challenge assumptions.
But we should also recognise the extraordinary opportunities ahead.
The more intelligence we can develop responsibly, the more difficult problems we may eventually be able to solve.
Diseases that can be prevented. Infrastructure that can be secured. Scientific questions we haven’t yet learned how to answer.
And perhaps, one day, a civilisation extending far beyond our own planet.
That’s a future worth building.
Have a wonderful Monday, everyone!
Never doom, only bloom.
Melissa Kampers