The AnewZ Opinion section provides a platform for independent voices to share expert perspectives on global and regional issues. The views expressed are solely those of the authors and do not represent the official position of AnewZ
In June 2026, Anthropic released Claude Fable 5, then the most powerful publicly available artificial intelligence model. Within days, researchers bypassed its safeguards, highlighting a growing reality: AI powerful enough to defend can also be used to attack.
In June 2026, Anthropic released Claude Fable 5, the most powerful publicly available artificial intelligence model at the time. Within days, Amazon researchers coaxed the model past its safeguards, effectively jailbreaking it using a prompt-based bypass to identify and exploit software vulnerabilities. The U.S. government responded with export controls, resulting in the model going dark for 19 days.
Fable 5 is now available again with a new safety classifier. OpenAI chose to preview its GPT-5.6 model to a tiny, U.S. government-approved group before releasing it publicly because of the same concerns.
A system smart enough to help defenders is also smart enough to help attackers.
So what does this tell us? It tells us that the most sophisticated AI companies, with hundreds of billions of dollars in investment and access to the world's leading talent, could not guarantee that their guardrails would withstand determined attempts to bypass them.
While every model is released with extensive protections, including inbuilt refusal training, content filters, monitoring and even a constitutional framework, they have still been defeated by determined users armed with nothing more than expertly crafted prompts. If this is true for software vulnerabilities, it is even more true - and far more consequential - for one of the defining issues of our time: disinformation.
Journalism sits at the centre of the automation paradox. The technology that is the most powerful tool our profession has ever witnessed is simultaneously the most powerful weapon pointed at it.
As a tool, AI has already demonstrated extraordinary capabilities. The cost of machine intelligence is collapsing faster than most of us can keep pace with. What cost more than $20 per million tokens in late 2022 now costs as little as $0.40, with much of that compression driven by Chinese models.
DeepSeek now offers high-performing models comparable to frontier systems at a cost of as little as $0.14 per million tokens. Chinese AI companies continue to slash prices, with the gap between American and Chinese pricing reaching as much as thirtyfold.
For journalism, this unlocks unprecedented capabilities. Tasks that once required a newsroom of 30 to 40 people — including transcription, translation, data visualisation, video generation, monitoring and drafting — can now be handled by a team of just two or three people in the Global South. AI tools that cost little more than two coffees in the West are powering newsrooms in Islamabad, Baku and Dhaka. I run an autonomous AI pipeline myself and have experienced first-hand how dramatically these tools improve newsroom efficiency. AI has democratised the production of journalism as never before.
The same collapse in AI costs also benefits those producing disinformation. Fake news stories that appear authentic have never been easier to create. Fake images, cloned voices and deepfake videos cost almost nothing to produce and can be generated in seconds. Even text prompts are no longer necessary, with spoken commands delivering similar results. The greatest danger, however, comes not from a single piece of misinformation but from the relentless flood of AI-generated content that increasingly dominates social media, shaping public perceptions every moment.
Machine-generated disinformation is easy to create and can be produced and distributed at virtually unlimited scale. Real journalism, even when supported by AI, cannot match that speed because human verification remains essential. As a result, genuine reporting risks being drowned out by noise while audiences become increasingly uncertain about what to trust. Consequently, public confidence in journalism continues to erode.
It could be argued that the same technology that has industrialised misinformation has also provided journalism with its most powerful verification tool. AI systems can cross-check claims at machine speed, flag synthetic content and compare information against millions of sources in the time it takes a human to review only a handful.
There is truth in that argument. News organisations are already using AI for fact-checking and verification. Yet verification still requires human judgement. A machine can verify a claim, but only a person can take responsibility for publishing it.
AI is already showing journalism two very different futures. On one hand, news organisations are deploying transparent, supervised automation that improves efficiency. On the other, some outlets have been caught publishing error-riddled AI-generated content under fake bylines, severely damaging public trust in the wider news ecosystem.
This is the crux of the argument: AI literacy is no longer optional. It has become a core component of journalism education and professional training - as important as interviewing skills and as non-negotiable as fact-checking.
Every working journalist in 2026 must understand how these models operate, what they can and cannot do, how they can be manipulated, and how they can be used responsibly for verification and fact-checking.
A new generation of journalists enters the profession every year, yet many still lack the ability to distinguish authentic content from synthetic material. Most journalism schools and training programmes have not kept pace with the speed of AI innovation.
The question is no longer whether machines can write the news. They can - and sometimes they can do so better than humans, as my own newsroom has experienced. The real question is whether journalists can use AI more effectively as a tool than those spreading disinformation can use it as a weapon.
To meet these global challenges, the journalism community must come together to establish shared standards for AI use, with AI literacy forming the cornerstone of any future strategy.
Hamza Azhar Salam is the Editor of The Pakistan Daily and a recent Journalist in Residence at the Stigler Center, University of Chicago, Booth School of Business. He writes columns for Haaretz and is a correspondent for France 24 in Pakistan. X: @HamzaAzhrSalam
An agreement between Iran and the U.S. to reopen the Strait of Hormuz is expected "soon," a U.S. official has said. The deal would result in the resum...
A state of emergency was declared in the district of Summerland, British Columbia, early on Saturday as a fast-moving wildfire swept through western C...
Typhoon Dolphin lashed Japan’s southern Okinawa prefecture on Saturday, injuring six people, disrupting flights and leaving more than 50,000 buildin...
An agreement between Iran and the U.S. to reopen the Strait of Hormuz is expected "soon," a U.S. official has said. The deal would result in the resumption of commercial shipping through the sea passage and the lifting of an American naval blockade on Iranian ports.
Start your day informed with AnewZ Morning Brief. Here are the top news stories for the 8th of August, covering the latest developments.
U.S. President Donald Trump said negotiations to reopen the Strait of Hormuz are progressing and suggested an agreement with Iran could be reached “soon”.
U.S. President Donald Trump may have to accept concessions he previously opposed if he wants to secure a deal with Iran to reopen the Strait of Hormuz, according to analysts, as negotiators move closer to a temporary agreement.
A U.S. judge has ordered Meta to pay an additional $567 million over claims that its platforms failed to protect children from online harm.
Europe's political centre is under growing pressure as migration, living costs and declining trust in mainstream parties reshape the continent. Berlin's 20 September state election may offer another indication that Germany's political landscape is entering a new phase.
The next political shock in Georgia may not weaken the ruling system. It may strengthen it.
Djibouti is a tiny country on the Horn of Africa with no significant natural resources, but it hosts multiple major foreign military bases located in close proximity. Why is that?
Oil shaped the geopolitics of the 20th century. People may shape much of the geopolitics of the 21st.
Azerbaijan's entry into the regional consultative format gives Central Asia reach across the Caspian. But turning geography into power will require more than summits and corridors.
You can download the AnewZ application from Play Store and the App Store.
What is your opinion on this topic?
Leave the first comment